A diagnosis of CD was histologically made in 17 (38.6%) of 44 enrolled patients, while the remaining 27 (61.4%) had a normal villous pattern. Among those diagnosed with CD, 7 (41%) showed endoscopic features of disease during SE.
After switching to NBI-ME, a diagnosis of mucosal abnormalities was made in 19 patients. Among these, 7, which were the same as those found at SE, had a complete absent duodenal villous pattern, while 12 displayed an abnormal duodenal villous pattern. Two cases showing abnormal villous patterns at NBI-ME were subsequently classified as negative for CD at histology (). NBI-ME identified patchy areas of partial villous atrophy in 12 patients with a sensitivity, specificity, PPV, and NPV of 100%, 93%, 89%, and 100%, respectively. The corresponding values for SE were 41%, 100%, 100%, and 73%, respectively (). The area under the ROC curve for NBI-ME was 0.978 (P = 0.0005), indicating an excellent agreement with the histological results. The area under the ROC was 0.706 for SE (P = 0.023) (). The overall agreement between NBI-ME and histology was significantly higher when compared with SE and histology (kappa score: 0.90 versus 0.46; P = 0.001) in determining CD (). Among 27 subjects without CD, the diagnoses were functional dyspepsia, irritable bowel syndrome, or overlap syndromes.
Histological diagnosis and corresponding endoscopic findings.
Diagnostic accuracy of SE and NBI with ME.
Receiver operating characteristic curves of NBI-ME and SE for diagnosing CD.
Overall agreement (k-Cohen coefficient) between endoscopic findings and histology reports.
The mean additional time for single NBI-ME procedure was 4 minutes and 30 seconds (±1 minute).
The first report on the use of ME with chromoendoscopy to highlight duodenal villous pattern was published by Siegel et al. [18
] demonstrating an increased detection rate of focal villous atrophy as well as partial villous atrophy compared with SE in patients with malabsorption. However, some limitations of this technique related to difficulty in achieving a complete and an even coating of the mucosal surface with the dye [12
], increased cost and procedure time, and the lack of visualization of the vascular pattern have prevented the widespread use of vital dye staining chromoendoscopy techniques. Recently a new technology called “Fuji Intelligent Color Enhancement,” a virtual chromoendoscopy, was introduced to enhance the contrast of the mucosal surface without the use of dyes, through the ability to select better spectral images decomposed from ordinary endoscopic images [19
]. Cammarota et al. [12
] published an original open, prospective, single-centre trial on the potential of a similar system, the optimal band imaging (OBI), for predicting the duodenal villous morphologic characteristics in patients with suspected CD. The authors concluded that the OBI system, in association with ME, allows clear visualization of the duodenal pattern, with a potential role in optimizing the diagnostic accuracy of endoscopy in CD.
In the present study we show that the NBI system, associated with ME, has provided superior performance than conventional endoscopy in detecting mucosal abnormalities on otherwise normal appearing duodenal mucosa. We found high sensitivity and specificity values with an overall agreement between NBI-ME and histology significantly higher when compared with SE and histology in determining duodenal villous pattern features. Moreover, the ROC curve analysis demonstrated that the NBI-ME performance was greater than SE, showing an excellent agreement with the histological results.
The NBI system consists of a sequential electronic endoscope system that can select better spectral images using particular luminous bands, thus enabling to filter incidence light resulting in some kind of “coloration without coloring.” Since the gastrointestinal tract is mainly composed of blood vessels and mucosa, narrow band illumination, which is strongly absorbed by hemoglobin and penetrates only the surface of tissues, is ideal for enhancing the contrast between the two.
However, our results have some limitations. First, we did not assess the inter- or intraobserver reliability of the mucosal patterns. Consequently, although duodenal villous classified patterns were normal, abnormal, or absent, we have not evaluated the correlation between partial or marked villous atrophy with histological score. In fact, our main objective was to establish the presence or absence of CD. A previous report by Badreldin et al. [10
] on the potential role of zoom endoscopy for the diagnosis of the various degrees of villous atrophy showed a sensitivity of 90.7% and specificity of 63%. In their study the main disagreement between zoom endoscopy and histopathology was the distinction between normal tall villi and morphologically normal but shortened villi, which depends on the assessment of villus height. More recently another study, using a simplified classification, demonstrated the feasibility of using NBI-ME for the detection of villous atrophy in patients presenting with suspected CD [20
Second, among the criticisms that have been raised for the use of NBI-ME was the slightly increased procedure time, including conscious sedation, for routine application. We observed an additional mean procedure time of 4 minutes and 30 seconds.
Third, as reported by other authors [12
], all patients who underwent UGI endoscopy had clinical history of malabsorption or serologic suspicion for CD, thus having a high pretest probability for duodenal abnormalities.
Fourth, we have found two false-positive cases at NBI-ME which reduced the specificity. It is possible that this approach could lead to an overestimation of the findings, and the results could be influenced by the high pretest probability for duodenal disease. It is widely known that the prevalence of CD in open access endoscopy is likely to be underestimated, with missed diagnose, ranging from less than 1% to 16% [21
]. The ideal diagnostic technique approach for CD should increase accuracy and sensitivity and be easy to perform, cost saving, and repeatable, and not be time consuming [22
Furthermore, other endoscopic options, such as confocal endomicroscopy [23
] and optical coherence tomography [24
], that were studied in patients with suspected CD achieved good results. However, these techniques are hampered by technical problems. In particular, difficulty of image acquisition/stability, with potential distortion and artifacts, and a long learning curve would restrict their wide use on the basis of the local availability of equipment and expertise. Moreover, according to Fedeli and colleagues [25
], the combination of two or more simple new endoscopic approaches, such as OBI together with ME, during water-immersion would obtain outstanding images of the villous pattern.
Using these new endoscopic techniques, the same authors have proposed an algorithm to minimize the need for duodenal biopsy in patients with suspected CD in particular for those with total villous atrophy [22
], or in such circumstances that could involve patients who are on anticoagulation therapy and that cannot be safely interrupted [26
]. As suggested in a correspondence [27
], we believe that it is not possible, at present, to avoid biopsies in CD, both for the initial diagnosis and for followup. In fact it is not possible to make a precise differential diagnosis of Giardia lamblia
infection or Crohn's disease (where there could also be changes of serological markers) [28
] and eosinophilic jejunitis or HIV enteropathy [29
]; furthermore, the histological evaluation is fundamental because it allows verifying improvement of duodenal lesions after gluten-free diet or the absence of mucosal recovery that needs an analysis of molecular markers in the suspicion of a T-cell dysplasia lymphoma [30