Introduction
The magnifying endoscopy (ME) with narrow-band imaging (NBI), which can visualize the microvascular (MV) architecture and the microsurface (MS) structure showed the acceptable diagnostic ability for early gastric cancer (EGC) even using second-generation NBI. The latest endoscopic system EVIS X1 (Olympus, Tokyo, Japan) includes third-generation (3G) NBI with more brightness and clearness, however, the diagnostic performance of ME with 3G-NBI for EGC was unknown.
Aims & Methods
The aim of this post hoc analysis was to investigate the diagnostic performance of ME with 3G-NBI for EGC using the data from the 3G detection trial, which was the randomized three-arm phase II trial comparing white light imaging (WLI), 3G-NBI, and texture and color enhancement imaging (TXI) for detecting gastric neoplasms (GN). The major inclusion criteria of 3G detection trial were as follows: 1) scheduled surveillance endoscopy after endoscopic resection for GN or endoscopic resection, chemotherapy, or radiotherapy for esophageal cancer (EC), or 2) scheduled preoperative endoscopy for known GN or EC. The EVIS X1 system and a high-definition gastroscope with an optical zoom (GIF-XZ1200) were used. Patients were randomly assigned to the following arms: WLI followed by WLI, 3G-NBI followed by WLI, and TXI followed by WLI. All suspected GN lesions were diagnosed as EGC or non-EGC using ME with 3G-NBI according to ME simple diagnostic algorithm of EGC (MESDA-G), which use the presence of irregular MV and/or MS patterns within the demarcation line, and its confidence level was recorded. Pathological diagnoses were made based on biopsied tissue or final results of resected specimens by expert pathologists at each institution. This post hoc analysis compared the diagnosis of ME with 3G-NBI for EGC with the pathological diagnosis. The primary endpoint was the diagnostic performance of ME with 3G-NBI for EGC, and it also evaluated according to confidence level (high/low), macroscopic type (elevated/flat or depressed), and lesion size (<10 mm/≥10 mm).
Results
Of 901 patients enrolled in the 3G detection trial, a total of 228 suspected GN lesions in 187 patients were analyzed. The median size of the lesions was 5.0 mm (range: 1.0-40.0). In the macroscopic type, 43 (18.9%) lesions were elevated and 185 (81.1%) were flat or depressed. Using ME with 3G-NBI, 62 lesions (27.2%) were diagnosed as EGC, of which 27 were diagnosed with high confidence. And, 166 (72.8%) were diagnosed as non-EGC, of which 91 were diagnosed with high confidence. Finally, 61 were pathologically diagnosed as EGC and other 167 were pathologically diagnosed as non-EGC. The overall diagnostic performance of ME with 3G-NBI for EGC was, sensitivity of 70.5% (43/61), specificity of 88.6% (148/167), accuracy of 83.8% (191/228), positive predictive value (PPV) of 69.6% (43/62) and negative predictive value (NPV) of 89.2% (148/166), respectively. For the lesions diagnosed with high and low confidence, sensitivity was 78.1% (25/32) and 62.1% (18/29), specificity was 97.7% (84/86) and 79.0% (64/81) (p<0.001), accuracy was 92.4% (109/118) and 74.6% (82/110) (p<0.001), respectively. For the elevated and flat or depressed lesions, sensitivity was 84.6% (11/13) and 66.7% (32/48) and specificity was 96.7% (29/30) and 86.9% (119/137), respectively. For the lesions with <10 mm and ≥10 mm in size, sensitivity was 65.7% (23/35) and 76.9% (20/26) and specificity was 88.6% (116/131) and 88.9% (32/36), respectively.
Conclusion
The diagnostic performance of ME with 3G-NBI for EGC was acceptable, especially in the lesions diagnosed with high confidence.
