Diffusion-weighted Magnetic Resonance Enterography for Evaluating Bowel Inflammation in Crohn's Disease: A Systematic Review and Meta-analysis

Background: To systematically determine the performance of diffusion-weighted imaging magnetic resonance enterography (DWI-MRE) for evaluating bowel inflammation in Crohn's disease and sources of heterogeneity between reported results. Methods: We identified research studies that investigated DWI-MRE to diagnose bowel inflammation (present versus absent) or to assess bowel inflammatory severity in Crohn's disease by performing a systematic search of PubMed MEDLINE and EMBASE (until March 31, 2015). Study quality was assessed using QUADAS-2. For studies reporting dichotomous diagnosis of bowel inflammation, study heterogeneity and threshold effect were analyzed, summary sensitivity and specificity were estimated, and meta-regression analysis was performed to further explore study heterogeneity. For studies reporting assessment of inflammatory severity, a qualitative summary was performed. Results: Of 159 articles screened, we found 12 studies (1515 bowel segments) reporting a diagnosis of bowel inflammation and 6 studies (1066 bowel segments) reporting assessment of inflammatory severity. The summary sensitivity and specificity were 92.9% (95% CI, 85.8%-96.6%; I-2 = 87.9%) and 91% (95% CI, 79.7%-96.3%; I-2 = 95.1%), respectively. Sensitivity and false-positive rate were inversely correlated (r = -0.650; P = 0.022). Lack of blinding to contrast-enhanced MRE when interpreting DWI-MRE (P = 0.01) and use of contrast-enhanced MRE as a reference standard (P < 0.01) in some studies were significant factors for study heterogeneity and likely caused overestimation of DWI-MRE accuracy. There was rather clear correlation between diffusion-related parameters and bowel inflammation severity, although the strengths were heterogeneous (correlation coefficient, 0.39-0.98). Conclusions: DWI-MRE accuracy was very heterogeneous between studies and was likely overestimated in some studies. Despite rather clear correlation between diffusion-related parameters and bowel inflammatory severity, its strength was variable.