NEAR-FOCUS NBI CLASSIFICATION OF VILLOUS ATROPHY IN SUSPECTED COELIAC DISEASE: INTERNATIONAL DEVELOPMENT AND VALIDATION

Aims:

To develop a reliable and validated endoscopic classification of villous atrophy in suspected coeliac disease using near focus Narrow Band Imaging (NF-NBI).

Methods:

Patients with symptoms/investigations warranting duodenal biopsy were prospectively recruited between September 2017 to August 2018. Six paired NF-white light (NF-WLE) and NF-NBI images with biopsy (2 from the first part of the duodenum,4 from the second) were obtained from each patient. Histopathology grading used Marsh-Oberhuber classification (M-O). Images were reviewed for quality and biopsy orientation. Separate images were used for development of the classification, training and validation steps. A modified Delphi process was performed on images and video recordings by 3 endoscopists to define NF-NBI characteristics (included if kappa> 0.6). 10 blinded endoscopists (3 expert, 7 non-expert) underwent a short training module on the proposed NBI classification and evaluated paired (NF-WLE/NF-NBI) images.

Results:

100 consecutive patients were recruited and n = 97 completed the study (66F, 51.2+/-17.3yrs). TTG positive n = 17/88. M-O VA (3a/3b/3c): n = 22. After image quality and biopsy orientation review; 573 paired images remained (M-O 0/1/2: n = 470;VA n = 103). 510 paired images developed the classification with modified Delphi; Villous shape, vascular discrimination, crypt phenotype. 10 endoscopists evaluated 50 paired images each (500 paired total observations made for validation). Sensitivity, specificity, NPV and accuracy of NF-NBI for the diagnosis of VA (Subtotal/total atrophy) using this classification:97.9 (95.2 – 99.3); 86.2 (81.4 – 90.1);97.8 (95.0 – 99.1);91.8 (89.04 – 94.05) respectively. Mean of difference in confidence using NF-NBI vs. WLE was 0.61 (0.51 – 0.71), p < 0.0001: The classification was further validated in histopathologically proven duodenitis (n = 15) images with no features of VA using the proposed classification.

Conclusions:

A novel NF-NBI classification for VA had been validated to reliably diagnose VA in suspected CD amongst both expert/non-expert endoscopists using readily available equipment and required only short training supporting translation to wider practice.