Diagnostic performance of light-induced fluorescence endoscopy for gastric neoplasms

Malignant tumors generate autofluorescent patterns that differ from those of normal tissue. However, whether autofluorescent diagnosis could be genuinely useful in screening for gastric neoplasms has not been well investigated in clinical practice. Accordingly, we retrospectively studied our experience with this diagnostic technique for various gastric lesions and assessed its diagnostic utility.

Autofluorescence diagnosis of 109 gastric lesions in 79 patients was done, without knowledge of the diagnosis by conventional white light endoscopy, retrospectively and independently by three endoscopists with 6 years', two years' and no experience of the technique. After examination of the interobserver bias in the assessment of autofluorescent pseudocolor in light-induced fluorescence endoscopy (LIFE), the relationship between pseudocolor and characteristics of gastric lesions (including histology, macroscopic type, and depth of invasion) were investigated.

The kappa statistic for agreement in pseudocolor diagnosis between the three endoscopists was 0.71. The assessment of pseudocolor by all of the observers was in agreement in 67 of the total of 109 lesions (61.5 %). Experience with the LIFE technique did not improve the accuracy of pseudocolor determination. All of the cancers, 87.5 % of the adenomas, and 50.9 % of the benign lesions were recognized as having an abnormal autofluorescent image. None of the gastric cancers and 49.1 % of the benign lesions were evaluated as having a normal autofluorescence image. The histopathological and macroscopic types of tumors and their depths of invasion were not reflected in the autofluorescence diagnosis.

LIFE provided a sensitivity of 96.4 % and specificity of 49.1 %, suggesting that this technique has limited clinical utility, regardless of the merits of acceptable interobserver bias and lack of necessity for experience with this technique.