Diffusion-weighted imaging for evaluating inflammatory activity in Crohn’s disease: comparison with histopathology, conventional MRI activity scores, and faecal calprotectin

This study aimed to evaluate the accuracy of subjective grading of DWI signal and objective measurement of ADC in grading of Crohn’s disease activity using a range of segmental and global standards of reference.

Our results were variable, with some data supporting the use of qualitative rather than quantitative DWI grading, but little evidence of utility of either against a segmental histopathological reference.

Although assessment of disease activity is vital in management of Crohn’s disease, there is no single accepted standard of reference with clinical scores, endoscopic, pathological, and biochemical markers used in clinical practice. In the present study, we used 3 reference standards: (1) faecal calprotectin, a biochemical marker of global gut inflammation; (2) a validated MRI global and segmental activity score [8]; and (3) histopathological analysis of endoscopic terminal ileal biopsy.

We found faecal calprotectin levels were significantly greater (reflecting greater disease activity) when enteric DWI was abnormal. The findings were similar when using MEGS as the standard of reference. The presence of any abnormal DWI signal had 83% sensitivity for active disease based on faecal calprotectin. However, specificity was very low at just 52%. We tested the theoretical addition of DWI to conventional MRI activity grading using MEGS and found a marginal benefit in terms of detecting active disease based on the calprotectin reference (increasing sensitivity from 88% to 91%). Our data are broadly similar to that of Kim et al, who also reported a small improvement in sensitivity for bowel inflammation by adding DWI to conventional sequences, at the expense of reduced specificity [22]. These data suggest DWI cannot replace conventional imaging sequences and is at best an adjunct to be used in combination. Although recent data suggest DWI may be able to replace contrast-enhanced imaging in standard protocols [5], it is clear that conventional MRI sequences are vital in maintaining specificity.

We did however find evidence that the longitudinal extent of DWI abnormality in the small bowel and colon was related to calprotectin level, i.e., in general, the more widespread the DWI abnormality, the greater the calprotectin level and inflammatory disease burden. A rapid review of DWI images noting the severity and extent of abnormal DWI signal may therefore provide a convenient and efficient method of assessing overall activity in patients with already known disease on conventional sequences. Such an approach could be useful in assessing treatment effect by rapidly comparing pre- and post-DWI imaging, for example.

To our knowledge, this is the first work to compare DWI with calprotectin. One advantage of calprotectin is that it better reflects total inflammatory burden than segmental references such as endoscopy which can assess at most the colon and terminal ileum. It does have limitations however. It may better reflect colonic rather than small-bowel Crohn’s disease [23], and levels are not necessarily linearly related to activity. These levels may also not be increased in the presence of active disease, or conversely may be falsely increased in the absence of inflammation [24]. Many workers have therefore compared DWI with conventional MRI activity scores, which we also explored in the present study. Hordonneau et al, for example, found the subjective presence of abnormal DWI had high sensitivity for segmental disease activity defined by a Maria score >7 [13]. Our data also suggested a relationship between DWI and an MRI activity score [8], although not as strong as those reported by Hordonneau et al. An important strength of the present study is that readers were fully blinded to conventional MRI sequences when grading DWI signal. Such an approach better isolates the true diagnostic worth of DWI and limits the major risk of incorporation bias when MRI is used as a standard of reference for new sequences.

Our data regarding the utility of quantitative ADC measurement was relatively poor. Similar to previous work [13], we did find a negative correlation between ADC and a segmental MRI activity score, although the strength of this association was weak.

Against a histopathological score of activity, neither qualitative segmental DWI grades nor ADC measurement showed a statistically significant relationship to the level of inflammation. This is at odds to previous studies which have shown reasonable correlations between both ADC and subjective DWI grading and an endoscopic standard of reference [16, 22].

Part of the discrepancy could in part be due to the standard of reference employed. The histopathological score is a based on mucosal biopsy whereas ADC measurements are taken from the whole bowel wall. The pathological scoring system has however been employed with success in other imaging studies [1, 25]. Furthermore, endoscopic standards of reference also suffer from the same limitation of assessing only the bowel mucosa and not the full transmural disease extent. Of note, in a detailed imaging–pathological correlation study using surgical resection specimens, Tielbeck et al found no association between ADC and a full wall thickness histological score of inflammation [26] which concurs with our findings.

Quantitative DWI analysis, with calculation of ADC, is a complex task. Several post-processing steps are necessary. First of all, regions of interest must be placed on the diffusion image sequences. These are then either transferred to the post-processed ADC maps for ADC calculation, or (as was performed in this study) the ROI is propagated through the each DWI sequence and the ADC values calculated directly.

The measurement of ADC in bowel is often difficult due to the relatively thin bowel wall, given the typical image slice thickness (normal bowel wall 1–2 mm, typical DWI slice thickness 5–8 mm). Bowel peristalsis during image acquisition introduces further error. These errors can be limited by shortening image acquisition time (e.g., fewer signal averages), but at the expense of a reduction in signal to noise. It is perhaps not surprising that interobserver agreement for measurement of ADC was poor between 2 readers in the present study. Indeed, it could be argued that the apparent reported reduction in ADC in active disease is in part due to overestimating ADC in the thin wall of normal bowel or that affected by fibrotic disease. Wall thickness per se is a strong predictor of activity [8‐10, 27], and measurement of ADC is easier in thickened bowel.

There are several limitations in this study. We used both prospectively recruited and retrospectively identified cohorts, although this was deliberate given our wish to apply a variety of reference standards. Patient numbers in each group were reasonable but not informed by any power calculation. As noted above, there is no single accepted reference standard for activity. By using calprotectin, an MRI activity score, and histopathology, we attempted to capture the strengths of several alternatives but did not employ others such as endoscopy grading or surgical specimens. We acquired data on 1.5 T and 3 T MRI, although we found no difference in ADC measurements between these two platforms. As noted above, a strength of our methodology is that we blinded readers to all conventional sequences when evaluating DWI to truly isolate its diagnostic potential. In clinical practice, this is not how DWI is employed. Similarly, the study design utilized consensus reading of DWI data, which may limit the generalizability of these results. The MR protocol was originally designed to allow biexponential quantification of diffusion data and therefore used many B-factors. Although the IVIM model has been shown to be of value [12], this protocol increases scan time significantly and therefore is prone to image degradation from bowel movement.

In conclusion, qualitative (visual inspection) assessment of DWI is a reasonably sensitive tool for detection of bowel inflammation in Crohn’s disease based on a calprotectin (but not histopathological) standard of reference but has poor specificity and cannot fully replace conventional sequences. Its ability to grade the severity of inflammation is however limited. In those with known disease, rapid review of DWI could give a reasonable estimate of the overall active disease burden. The addition of DWI to conventional MR enterography sequences has a marginal benefit in increasing diagnostic accuracy. Quantitative ADC measurements are prone to poor interobserver agreement, have lower discriminatory ability for active disease, and are therefore not recommended.