Data from 33 consecutive patients with Crohn disease proven at histopathologic analysis were analyzed. These patients had taken part in a prospective single-center study comparing dynamic contrast-enhanced (DCE) MR enterography to ileocolonoscopy with CDEIS. The indication for ileocolonoscopy was clinical suspicion of relapsing Crohn disease. Exclusion criteria were age younger than 18 years, contraindications for MRI (including pacemakers, metallic implants, severe claustrophobia, and pregnancy), technical failure of a sequence, incomplete reference standard (CDEIS), and a negative diagnosis for Crohn disease. All patients had been recruited between February 2009 and November 2010 for assessment of Crohn disease activity. Furthermore, for all patients, ileocolonoscopy had been performed within a month of the MR enterography. The results of that study have been published previously [16].

Patient exclusion criteria for this study were non-diagnostic MR enterography image quality (i.e., the study was not of sufficient quality to determine disease activity, if present) as determined by one or more of the readers and an incomplete MR enterography scan protocol (i.e., not including T2-weighted single-shot fast spin-echo [FSE], T2-weighted fat-saturated single-shot FSE, or 3D T1-weighted contrast-enhanced sequences, which are mandatory for calculating the scoring systems). Per-segment exclusion criteria were resected bowel segments and insufficient distention or visibility (< 20% of the bowel adequately distended and visible) of a bowel segment, as determined by one of the readers.

The Crohn disease activity index (CDAI) score [17] and C-reactive protein levels were assessed in all patients. A CDAI score greater than 150 or a C-reactive protein level greater than 8 mg/L was considered as active disease.

For the previous study, ethical permission was obtained from the hospital medical ethics committee, and written informed consent was obtained from all patients. For the current study, informed consent was waived by the hospital medical ethics committee.

Finally, three of the 33 patients in the dataset of the prior study were excluded. Two observers assessed the quality of one MR enterography study (one patient) as nondiagnostic, and for two other MR enterography examinations, no T2-weighted fat saturation sequence was available (two patients). Thus, 30 patients (median age, 32 years; age range, 19–72 years; 21 women and nine men) were evaluated. The CDAI values showed that 47% (n = 14) of the patients had active disease. The C-reactive protein values showed that 50% (n = 15) of the patients had active disease. Baseline characteristics are shown in Table 1.

These 30 patients had 148 segments (in two patients only four segments were eligible after right hemicolectomy) of which five segments, all rectal, were excluded because of insufficient visibility, resulting in 143 evaluable segments. The remaining 143 segments were radiologically scored by the four observers.

The protocol of the study has been published previously [16]. Patients fasted 4 hours before the examination and drank 1600 mL of mannitol (2.5%; Osmitrol, Baxter) solution 1 hour before the scan. Supine images were acquired using a 3-T MRI unit (Intera, Philips Healthcare) with a 16-channel torso phased-array body coil. Axial and coronal T2-weighted single-shot FSE sequences with and without fat saturation were acquired, followed by a coronal 3D T1-weighted spoiled gradient-echo sequence with fat saturation. After these series, 20 mg of butylscopolamine bromide (Buscopan, Boehringer Ingelheim) was IV administered, and a DCE-MRI sequence with 0.1 mL/kg bodyweight of gadobutrol (1.0 mmol/mL; Gadovist, Bayer Schering Pharma) was obtained. Ten seconds after the start of the dynamic sequence, 0.1 mL/kg bodyweight of gadobutrol (1.0 mmol/mL) was injected IV by bolus injection (5 mL/s) through a 20-gauge IV catheter using an automated injection pump (Mallinckrodt Optistar, Liebel-Flarsheim). Injection of contrast medium was immediately followed by a bolus of 15 or 20 mL saline (5 mL/s), depending on the length of the contrast injection tube. The duration of the DCE-MRI sequence was 6 minutes. After these series, a second dose of 20 mg of butylscopolamine bromide was IV administered. Thereafter, contrast-enhanced axial and coronal 3D T1-weighted spoiled gradient-echo sequences with fat saturation were performed. All sequences were used for image analysis, except the DCE-MRI sequence.

Four readers from two tertiary centers in different countries with 18 years (700 MR enterography studies), 17 years (1100 MR enterography studies), 4 years (170 MR enterography studies), and 1 year (160 MR enterography studies), of experience in reading abdominal MRI evaluated the MRI scans using the axial and coronal T2-weighted single-shot FSE with and without fat saturation, coronal unenhanced, and axial and coronal contrast-enhanced 3D T1-weighted spoiled gradient-echo sequences (Table 2). All readers used a PACS (Impax 5.0, AGFA Healthcare, Agfa-Gevaert) workstation. All readers were unaware of the findings at the initial reading and the findings from ileocolonoscopy but were aware of patients' surgical history. The small bowel and the colon were divided into five segments: terminal ileum, right colon (cecum plus ascending colon), transverse colon, left colon (descending colon plus sigmoid), and rectum, so there could be a direct segment comparison between MRI and the CDEIS.

Seventeen different MRI features (Table 3) were evaluated by all readers. Features were selected according to the MRI features described in the literature and used by most abdominal radiologists as identified in an international inventory, together with those used in the two published scoring systems [13, 15, 18]. The most affected part of the segment was chosen for scoring.

The following MRI features were used to calculate the MR index of activity: mural thickness in millimeters, relative contrast enhancement, and the presence of edema and ulcers. These features have been proven to be significantly correlated to the CDEIS. The MR index of activity was calculated using the following formula: (1.5 × wall thickness in millimeters) + (0.02 × relative contrast enhancement) + (5 × edema) + (10 × ulceration) [13].

For the overall CDMI score, the following four features—mural thickness, mural T2 signal, perimural T2 signal, and mural T1 enhancement—were scored on a scale of 0 to 3, resulting in a maximum score of 12 [15]. These four features were selected because they were found to be significantly correlated with disease activity according to an endoscopic biopsy acute inflammatory score. In addition, the sum of the scores for mural thickness, mural T2 signal, perimural T2 signal, and contrast enhancement showed the highest accuracy [15]. Furthermore, the following features were assessed: abscess, comb sign, enlarged (> 1 cm) lymph nodes, fistulas, lymph node enhancement, pattern of mural enhancement, pseudopolyps, and total length of the disease in each segment.

The MRI features with regard to lymph nodes were scored per patient; all other features were assessed per segment. The readers used the same method as described in detail in the articles about the MR index of activity and the CDMI score [13, 15]. For calculating the relative contrast enhancement involved, we used the formula as described in Rimola et al. [13] [(WSI contrast-enhanced – WSI unenhanced) / WSI unenhanced] × 100 × (SD noise unenhanced / SD noise contrast-enhanced). Here, WSI is the wall signal intensity, SD noise unenhanced corresponds to the average of three SD of the signal intensity measured outside of the body before gadolinium-based contrast agent injection, and SD noise contrast-enhanced corresponds to the SD of the same noise after gadolinium-based contrast agent administration [19]. Ulcerations (defined as deep depressions in the mucosal surface of a thickened segment) and the short-axis diameters of enlarged lymph nodes were assessed on contrast-enhanced 3D T1-weighted spoiled gradient-echo images with fat saturation.

Eight of the MRI features are common to the MR index of activity and CDMI score but were assessed using different definitions of abnormality according to the particular scoring system. Specifically, mural thickness was measured using an ordinal score (0–3) and as a continuous variable in millimeters using calipers on either single-shot FSE or spoiled gradient-echo sequences. T1 contrast enhancement was measured using an ordinal score (0–3) and using relative contrast enhancement [19]. Lymph nodes were assessed using an ordinal (0–3) score and a binominal score (yes/no). T2 signal was measured using an ordinal score (0–3) and a binominal score (edema; yes/no).

Colonoscopy was performed after standard bowel preparation by either a gastroenterologist or a senior resident in gastroenterology under direct supervision of a gastroenterologist, using a standard colonoscope. The performing endoscopist was aware of the patient's history but was blinded to the MR enterography results. Segments were excluded from the analysis if they could not be scored during ileocolonoscopy. One of two gastroenterologists experienced in endoscopy of inflammatory bowel disease assessed the CDEIS [1]. A segmental CDEIS was calculated using the variables of deep ulceration (no = 0, yes = 12), superficial ulceration (no = 0, yes = 6), surface involved by disease (0–10), and ulcerated surface (0–10) for each of five bowel segments (terminal ileum, right colon, transverse colon, left colon, and rectum). The reference standard has been described in detail elsewhere [16]. In six patients, the terminal ileum could not be assessed during ileocolonoscopy because of a stenosis; therefore, 137 segments were correlated to the segmental CDEIS scores.

The median time between colonoscopy and MR enterography was 7 days (interquartile range, 5–14 days). MRI and colonoscopy were not performed on the same day. The median CDEIS was 4.3 (interquartile range, 1.6–5.8).

Several multirater analyses were performed for all features individually and for the overall MR index of activity and CDMI score to assess the interobserver agreement. For all ordinal data, a weighted kappa coefficient was calculated per two raters and eventually was pooled. For the binomial data, a multirater kappa coefficient was used, which was also calculated per two raters and pooled. For continuous data, a multirater intraclass correlation coefficient was determined.

In addition, the scores of both of the most experienced abdominal radiologists were analyzed post hoc. This was done to evaluate whether experience would positively influence the reproducibility values. Both the kappa and intraclass correlation coefficient values interpretation was as follows: 0–0.20, poor; 0.21–0.40, fair; 0.41–0.60, moderate; 0.61–0.80, good; and 0.81–1.00, excellent [20].

For an overall correlation, we first calculated means of MR index of activity scores and CDMI scores per segment for all four observers and correlated these values with the CDEIS scores. Because the segmental scores were interpreted as continuous variables, the Spearman correlation was used to correlate the segmental CDEIS scores to the segmental MR index of activity and segmental CDMI scores. Correlation coefficient values were interpreted as follows: 0.0, not correlated; 0.2, weakly correlated; 0.5, moderately correlated; 0.8, strongly correlated; and 1.0, perfectly correlated. Statistical analysis was performed in Excel 2003 (Microsoft) using PASW statistics software (version 19, SPSS).

All MRI features showed a fair to good interobserver variability (Table 4). Wall thickness measured in millimeters reported highest agreement of 0.69 (95% CI, 0.62–0.75). In addition to wall thickness measured in millimeters, the presence of edema, the pattern of enhancement, and the length of the disease in each segment showed a good interobserver variability between all readers (Figs. 1–3).

The results for measuring wall thickness in millimeters for experienced radiologists were significantly better than the multirater data and showed an excellent interobserver variability of 0.87 (95% CI, 0.82–0.90). The incidence of abscesses, fistulas, and pseudopolyps was very low in our study. Therefore, only the overall agreement among all readers (0.96, 0.94, and 0.99) is reported. Good interobserver variability was reported for abscesses and fistulas between the most experienced readers (0.66 and 0.66). Furthermore, the comb sign and lymph node assessment improved from fair to moderate agreement (0.51 and 0.58, respectively) for the experienced readers, suggesting that prior experience with MR enterography is an advantage in the assessment of these features.

Reproducibility—The MR index of activity and CDMI scores showed good interobserver variability (0.74 and 0.78, respectively) in the assessment of Crohn disease activity for all readers (Table 5). There was a minimal increase in the reproducibility (0.80 and 0.81, respectively) for the MR index of activity and the CDMI scores between the two experienced readers only. For the segmental scores of the four observers, median MR index of activity was 4.90 (range, 1.22–35.62), and the median CDMI score was 0 (range, 0–12) (Fig. 4).

Correlation to the CDEIS—Both MR index of activity and CDMI scores correlated moderately with segmental CDEIS (r = 0.51 [95% CI, 0.38–0.63] and r = 0.59 [95% CI, 0.47–0.69], respectively).