Background: Accurate identification and staging of gastric intestinal metaplasia (GIM) during endoscopy is challenging, in particular in countries with a low incidence of gastric cancer (GC). Random biopsy histopathology, remains the current gold standard for GIM staging, although random biopsy sampling is poorly reproducible which hinders individual risk prediction and might lead to an underestimation of GIM. Virtual chromoendoscopy (VCE) has shown to be superior to white light endoscopy (WLE), achieving > 90% accuracy in countries with middle-high GC incidence. We evaluated clinical yield of an endoscopy-led risk stratification strategy combining VCE and targeted GIM biopsies in low GC incidence settings.
Methods: A prospective study was conducted in the UK and the Netherlands, including patients with a known premalignant gastric lesion (PGL). All participants underwent two endoscopies; standard WLE with random biopsies at first endoscopy, and VCE with targeted biopsies by an expert endoscopist at second endoscopy at least six months later. During second endoscopy, the endoscopist systematically graded GIM appearance for each gastric location with targeted biopsies of either suspected GIM and/or non-GIM. Diagnostic accuracy was determined by the total number of targeted biopsies confirming GIM presence or absence divided by all biopsies taken. Endoscopists and pathologists were blinded to prior histopathology results. For final comparison of current standard practice to an endoscopy-led approach, diagnostic yield of WLE with random biopsies was compared to diagnostic yield of VCE with targeted biopsies.
Results: A total of 120 patients were included (mean age 62.7, SD 13.3, female: 53.3%). VCE combined with targeted biopsies identified GIM in 109 (90.8%) patients, 34% with an OLGIM stage >2 and 34% with extensive GIM. Diagnostic accuracy (95% CI) of VCE for GIM was 80.9% (77.5 – 83.9). Location-specific accuracy was 77.8% (73.2 – 82.0) for antrum and 85.4% (80.4 – 89.6) for body. The area under the curve (AUC) of the ROC curve for identifying extensive GIM was 0.787 (0.702 - 0.867). In 86 patients, diagnostic yield of WLE with random biopsies was compared to VCE with targeted biopsies. During WLE, GIM was found in 88.4% and extensive GIM in 36%, compared to 93% and 37.2% for VCE. VCE increased OLGIM stage in 34.9% of patients, with 46.5% and 18.6% showing the same or decreased OLGIM stage, respectively.
Conclusion: In low GC incidence countries, our study demonstrates that the accuracy of VCE for GIM is 80.9%. VCE combined with targeted biopsies results in a high GIM detection rate, with an AUC 0.787 for identifying extensive GIM. Although the accuracy is lower than diagnostic accuracy reported in middle and high GC incidence countries, our findings support endoscopy-led risk stratification in patients with PGL, even in low incidence countries.

Ernst Kuipers

Erasmus University Medical Center

Appearances