Comparative evaluation of new and conventional classifications of magnifying endoscopy with narrow band imaging for invasion depth of superficial esophageal squamous cell carcinoma

2017 
: A new classification of magnifying endoscopy with narrow band imaging (ME-NBI) for diagnosing and staging superficial esophageal squamous cell carcinoma (SESCC) was proposed by the Japan Esophageal Society in 2011. This study aimed to compare the new classification with the conventional classifications (Inoue's classification and Arima's classification). This was a prospective analysis of data from a single cancer center involving 151 consecutive patients with 156 SESCCs that were endoscopically or surgically resected. Initially, only ME-NBI images were selected and reviewed independently by three experienced endoscopists. White light imaging (WLI) was then evaluated separately after an interval. The diagnostic performance of each classification and interobserver agreement were assessed, and the WLI findings that affect the diagnosis by the new classification were identified. The specificity for classifying invasive depth as epithelium (EP)/lamina propria mucosae (LPM) confined was higher with the new classification than with Inoue's classification (0.512 vs. 0.349; P = 0.02) and Arima's classification (0.512 vs. 0.279; P < 0.01). However, the sensitivity was lower (0.902 vs. 1.000; P < 0.01) compared with Arima's classification. The concordance rates of three evaluators (κ values) were 0.52 for the new classification, 0.50 for Inoue's classification, and 0.23 for Arima's classification. On multivariate analysis, thickness on WLI independently affected the accuracy of diagnosis with the new classification (OR 3.23; 95%CI, 1.30-8.03). The new classification is superior to conventional classifications with respect to specificity for diagnosing SESCC with depth EP/LPM. Thickness on WLI was a factor negatively affecting the diagnostic performance of the new classification.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    10
    References
    9
    Citations
    NaN
    KQI
    []