Letter: accuracy of magnetic resonance index of activity score to predict response to biologics in Crohn's disease—authors' reply

We thank Dr Debnath & Prof Rathi for their letter¹ raising some points on our recent publication focused on the value of magnetic resonance enterography (MRE) to predict treatment response to TNF inhibitors in patients with Crohn's disease (CD).² Our study provides novel insights into the potential role of MRE in the management of CD, particularly after the observation from our own group that baseline ileocolonoscopy findings fail to predict response to TNF inhibitors.³

In our prospective study, we included in the multivariable analysis a number of features detected by conventional imaging sequences (thickness, oedema, enhancement, ulcers, fat hypertrophy among others), qualitative and quantitative evaluation of functional sequences (diffusion weighted imaging) and also the MaRIA score that is a validated MRE-based activity index against endoscopy with a high reproducibility and interobserver agreement.^{4, 5} We preferred the use of this well characterised index over other non-validated indices to better understand the results and its comprehensive implementation in further studies.

The presence of peri-enteric fat hypertrophy, lesion location (ileal vs colonic) and, in the case of ileal lesions, the severity measured by MaRIA score represent the main predictive MRE features of response to TNF-inhibitor therapy. However, the concept of inflammation burden deserves further investigation.

The endpoints we have applied in our study did not differ from the original MaRIA score.⁴ A segmental MaRIA score ≥11 has consistently been shown to have a highly predictive value for the presence of ulcerations on endoscopy, both in our studies^{5, 6} and in studies from other groups.⁷ In addition, and to avoid optimistic results on the achievement of ulcer healing (MaRIA <11 on those segments with MaRIA ≥11 on the pre-treatment MRE), we added a sensitivity analysis with a more stringent endpoint defined as a MaRIA <11 plus a reduction of 5 points relative to baseline. The similarities in the diagnostic accuracy between the original model of prediction with those obtained on the sensitivity analysis add robustness to our findings, and open the possibility for its use in future clinical research investigations.

Diffusion weighted imaging (DWI) was not shown to be predictive of response in our study. We regret not sharing the enthusiasm on the use of DWI as a critical piece in the assessment of inflammation in CD. Evidence shows that this sequence has important limitations including low specificity to detect inflammation, thereby leading to overdiagnosis,⁸ inconsistent data depending on the technical aspects such as bowel distension, ⁹ and insufficient diagnostic performance for monitoring the disease.¹⁰ Further evidence is needed before the role of DWI in clinical practice and research can be fully established.

We hope our clarifications help in understanding the methodology, results and interpretation of our study.