In vivo histologic diagnosis of gastric intestinal metaplasia (GIM) and gastric cancer (GC) can be achieved by confocal laser endomicroscopy (CLE). This review describes the endomicroscopic features of GIM and GC and reviews their clinical applications. Differentiation of phenotypes of GIM and GC by using CLE is also discussed.
Gastric intestinal metaplasia (GIM) is regarded as a precancerous lesion with the risk of developing into intestinal-type gastric cancer (GC) ; hence, endoscopic diagnosis of GIM is valuable for patients undergoing surveillance endoscopy. Biopsy is normally required to confirm the diagnosis of GIM because diagnosis of GIM by using conventional endoscopy is unreliable.
On the other hand, endoscopy is the current touchstone of GC diagnosis. With the development of endoscopy, endoscopists nowadays can predict not only the malignancy of a lesion but also its histologic phenotype. During the evolution of endoscopic diagnosis, knowledge and technology are inseparable. There is no doubt that as technology advances, endoscopists could give real-time histologic diagnosis.
Confocal laser endomicroscopy (CLE), with its optical biopsy utility aided by fluorescein contrast agents and the most significant magnification power at present, assists in the in vivo histologic diagnosis of GIM and GC. The aim of this review is to describe the endomicroscopic features of GIM and GC and review their clinical applications.
Staining agents in CLE
Fluorescein sodium is the most common contrast agent used in CLE, and intravenous fluorescein sodium alone is sufficient in the diagnosis of GIM. In the diagnosis of GC, tissue structure and microvessels could be well displayed by using fluorescein sodium. Although the outline of each cell is identifiable with fluorescein sodium, the intracellular architecture, particularly the nuclei, could not be displayed.
In the allergy test before the procedure, 1 mL of 2% fluorescein sodium should be given. For observation with CLE, another standard 5 mL of 10% fluorescein sodium is intravenously injected. Severe complications were not reported so far. Mild allergic reactions including pruritus and rash are rare and can be overcome by temporary antiallergic treatment. The most common side effects, yellow-stained skin and urine, usually resolve in 24 hours.
With specific binding to DNA, topical acriflavine stains nuclei specifically under CLE. The most frequently applied concentration of acriflavine is 0.02%. Too high concentration of acriflavine stains the whole cell, which makes nuclei unidentifiable, and too low concentration decreases image quality.
Although this agent was reported as having risks of carcinogenesis in animal experiments, evidence is still lacking for human beings.
Cresyl violet is another common staining agent that negatively stains the nuclei in CLE image. However, current evidence supports its application only in the colon and small intestine. Evidence of its application in stomach is not yet available.
Microscopy of GIM
The conventional diagnosis of GIM still depends on biopsy and histology, although advanced white light endoscopy could predict the histologic diagnosis more accurately than conventional endoscopy. The histologic characterization of GIM is essential for its in vivo CLE diagnosis.
The present classification of GIM consists of type I (complete) and types II and III (incomplete). Type I (complete GIM) appears similar to the normal small intestinal mucosa, with goblet cells and absorptive cells arranged beneath a brush border. In incomplete GIM, the typical enterocytes are replaced by intermediate cells lacking a well-developed brush border. Type II shows glandular distortion and consists of goblet cells and columnar mucous cells secreting sialomucins. Type III is characterized by glandular distortion and columnar mucous cells containing sulfomucins.
Type III is associated with higher risks of development into intestinal-type GC. Although some investigators suggested that the predictive value of GIM types be limited, others recommended that patients with type III GIM be closely followed up. Therefore, the diagnosis of GIM involves both identification and classification.
Endoscopy of GIM
There are no standard endoscopic features that distinguish GIM from the normal mucosa. Detection of GIM by using conventional endoscopy relies on random biopsy and histology. At present, the in vivo endoscopic identification of GIM is achieved only by using some advanced endoscopic technologies, such as chromoendoscopy, magnification endoscopy, narrow band imaging (NBI), and enhanced magnification endoscopy, alone or in combination. The endoscopic features of GIM are based on the assessment of gastric pit patterns and color changes.
The observation by Guelrud and colleagues using enhanced magnification endoscopy reveals 4 patterns of the gastric mucosa: (1) round pits, a characteristic pit pattern of regular and orderly arranged circular dots; (2) reticular, pits that are circular or oval and regular in shape and arrangement; (3) villous, no pits present but a fine villiform appearance, with the villi having a regular shape and arrangement; and (4) ridged, no pits present but cerebriform with regular shape and arrangement of the villi. The results of their observation suggest that type III and type IV correlate with the histologic identification of GIM. Although the sensitivity of enhanced magnification endoscopy for detection of GIM was good (94%), the specificity was poor (64%). The classification system was not applied to differentiate complete from incomplete GIM. In addition, the intraobserver and interobserver variabilities were not evaluated.
Some investigators suggested the combination of chromoendoscopy and magnification endoscopy to be applied in both identification and classification of GIM. Dinis-Ribeiro and colleagues classified the gastric mucosa into 3 groups based on 2 variables: pit patterns and change in color of the mucosa after application of methylene blue. Group I is defined as having regular pit pattern and no color change, which represents nonmetaplastic nondysplastic mucosa. Group II is defined as having regular pit pattern and blue color change, which represents GIM. Group III is defined as having neither a clear pit pattern nor a clear color change, which represents dysplastic mucosa. Group II is divided into 4 subgroups according to detailed pit patterns as follows: blue irregular marks (IIA), blue round and tubular pits (IIB), blue villi (IIC), and blue small pits (IID). The study revealed that IIA and IIB were more often associated with complete GIM and IIC and IID with incomplete GIM. For the diagnosis of GIM, this classification had an accuracy of 82%, a sensitivity of 76%, and a specificity of 87%. The overall accuracy of the diagnosis of complete and incomplete GIM was 84% and 82%, respectively. In reproducibility assessment, although interobserver and intraobserver agreements were substantial for classification by groups, the agreements were only moderate for classification by subgroups, such as types IIA to IID. Therefore, the diagnostic criteria are reliable for identification of GIM, but the classification of GIM based on them needs to be confirmed.
NBI is a novel endoscopic technology that is based on the principle of modifying the spectral characteristics of the illuminating light by narrowing the bandwidth of the optical filter in the light source. Some studies showed that the appearance of a light blue crest (LBC) in the mucosa is a distinctive endoscopic finding, suggesting an increased likelihood of detecting GIM. The LBC appears as bluish white patchy areas in NBI. Uedo and colleagues reported that identification of LBC had a sensitivity of 89%, a specificity of 93%, and an accuracy of 91% for the diagnosis of GIM. It was speculated that the appearance of the LBC was caused by the difference in the reflectance of the light at the surface of the ciliated tissue structure. Whether LBC could be applied in the classification of GIM was not reported, and the interobserver or intraobserver variability needs further evaluation.
In summary, the current endoscopic diagnosis of GIM relies on indirect characteristics such as pit patterns and color change (by staining agents or optical methods). These characteristics increase the accuracy of in vivo diagnosis of GIM. However, the direct evidences of GIM, such as presence of goblet cells, could not be obtained, and classification of GIM phenotypes by these technologies is still difficult. Although the combination of chromoendoscopy and magnification endoscopy could differentiate complete from incomplete GIM, with moderate interobserver agreement, the preparation and process of chromoendoscopy are complicated and time consuming. Furthermore, a recent report suggested that methylene blue chromoendoscopy induced questionable oxidative DNA damage.
Endomicroscopy of GIM
As a combination of endoscopy and in vivo microscopy, endomicroscopy should serve in the real-time identification and classification of GIM.
Identification of GIM by endomicroscopy
The endomicroscopic identification of GIM is based on both histology and endoscopy, the latter depends mostly on evaluation of gastric pit patterns. Therefore, the criteria for CLE diagnosis of gastric lesions, including GIM, begin with the classification of gastric pit patterns. According to the study by Zhang and colleagues , the gastric pit patterns are classified into 7 types:
Type A; round pits with round openings, representing normal mucosal with fundic glands.
Type B; noncontinuous, short, rodlike pits with short threadlike openings, representing corporal mucosa with chronic inflammation.
Type C is continuous, short, rodlike pits with slitlike opening, representing normal mucosa with pyloric glands.
Type D is elongated, tortuous, branchlike pits representing antral mucosa with chronic inflammation.
Type E is decreased number of prominently dilated pits representing chronic atrophy gastritis.
Type F represents GIM and has villuslike appearance, an interstitium in the center, and goblet cells.
Type G represents GC with normal pits disappearing, with appearance of atypical cells or glands.
Guo and colleagues refined the CLE imaging of GIM into 3 features: goblet cells, columnar absorptive cells and brush border, and villiform foveolar epithelium. GIM was determined if any of the 3 features were present. The diagnosis based on the 3 features has a sensitivity of 98.13% and a specificity of 95.33%. The kappa score for the correlation between CLE and histopathology was 0.94. Details of the 3 GIM features are as follows:
Goblet cells: large black cells with mucin
Columnar absorptive cells and brush border: more slender and brighter than columnar mucous cells of normal gastric mucosa, with a clear dark line at the surface of the epithelium
Villiform foveolar epithelium: a typical villouslike appearance different from the antral or corpus foveolae gastricae.
Among the 3 features, presence of goblet cells seems to be the most sensitive marker of GIM under CLE. Presence of goblet cells is also the well-established hallmark of GIM in conventional histology, because they are absent in normal gastric musoca. Bao and colleagues used two-photon fluorescence endomicroscopy to examine the mouse intestine, which has goblet cells, as a model of intestinal metaplasia. One-photon confocal fluorescence endomicroscopy and two-photon fluorescence endomicroscopy were used for 3-dimensional imaging of goblet cells. Both the techniques can 3-dimensionally view goblet cells in mouse large intestine and achieve an imaging depth of 176 μm.
Although with significant features, diagnosis of GIM by using CLE is still limited by the penetration depth, which is 250 μm at most. GIM in deeper layers of the mucosa could be missed by current CLE techniques.
Classification of GIM by endomicroscopy
In the study by Guo and colleagues, GIMs were classified as complete (type I) and incomplete (type II or type III) based on morphology and mucin staining with alcian blue (AB) with or without periodic acid–Schiff and high-iron diamine with or without AB, the mucin staining being applied in parallel sections for this purpose.
In the CLE images, GIM was also further classified as complete or incomplete based on the shape of goblet cells, the presence of absorptive cells or brush border, and the architecture of vessels and crypts as follows:
Complete: goblet cells interspersed among absorptive cells with or without brush borders; with regular crypts and capillaries
Incomplete: smaller numbers of goblet cells scattered among gastric type cells (mucous cells); without absorptive cells and brush border; with tortuous and branched crypts or irregular capillaries.
The sensitivity, specificity, positive predictive value, and negative predictive value of CLE for the diagnosis of complete GIM was 68.03%, 89.66%, 84.69%, and 76.92%, respectively, and those of incomplete GIM was 68.42%, 83.41%, 40.63%, and 94.09%, respectively. The kappa score for the agreement between CLE and histopathology was 0.67. Fig. 1 shows the CLE images of the subtypes of GIM and their corresponding histologic features.
Microscopy of GC
Several histologic classifications of GC, such as the World Health Organization, Ming, Lauren, Nakamura, and Goseki classifications, have been proposed. The Lauren classification presented in 1965 was widely accepted. This classification has advantages in that it distinguishes, by microscopic morphology alone, 2 main cancer pathogeneses that have clearly dissimilar clinical and epidemiologic entities (diffuse and intestinal subtypes). The intestinal-type GC is conventionally regarded as the well-differentiated type, and the diffuse-type GC as the poorly differentiated type. In general, intestinal-type tumors have a glandular pattern accompanied by papillary structure or dense components. The glandular pseudostratified epithelium consists of large pleomorphic cells with increased nuclear to cytoplasmic ratio. The polarization of columnar cells is usually preserved. On the contrary, diffuse-type GC is predominantly composed of atypical cells, without any forms of gland.
Similar to the diagnosis of GIM mentioned earlier, the diagnosis of GC by using CLE also includes identification and classification.
Endoscopy of GC
Because advanced GC is easily detected by using conventional endoscopy, the focus of endoscopic diagnosis of GC is on the detection of early GC (EGC). Advanced endoscopy for diagnosis of EGC also includes chromoendoscopy, magnification endoscopy, enhanced magnification endoscopy, and NBI. Given the complicated and time-consuming process of chromoendoscopy, combining it with conventional endoscopy is not practical for the in vivo diagnosis of EGC. Chromoendoscopy alone is still used only for detection of suspected lesions, whereas chromoendoscopy in combination with magnification endoscopy is generally required for further identification and classification of EGC. Some investigators suggested that the fine surface pattern of gastric carcinomas is often too subtle to be detected with a standard magnification endoscope, and adherent mucus or excessive dye makes magnification chromoendoscopy of the stomach difficult. So most of the recent studies were focused on magnification combined with NBI and enhanced magnification endoscopy. The endoscopic diagnosis is mainly based on assessment of surface pattern and microvascular alterations.
Tanaka and colleagues classified the surface pattern of gastric mucosa into 5 types by using enhanced magnification endoscopy:
Type I, small round pits of uniform size and shape
Type II, slitlike pits
Type III, gyrus and villous patterns
Type IV, irregular arrangements and sizes of pattern types
Type V, destructive patterns.
In a pilot study involving 47 patients with EGC or adenoma, the investigators found that types IV and V were more associated with EGC. To evaluate the efficacy of the classification for identifying EGC, they conducted conventional magnification endoscopy, magnification chromoendoscopy, and enhanced magnification endoscopy on 380 consecutive patients. The results showed that enhanced magnification endoscopy was significantly superior to conventional magnification endoscopy and magnification chromoendoscopy in detection of EGC. Classification of types IV and V strongly correlated with the presence of GC (sensitivity 100%, specificity 89.7%). However, the positive predictive value was only 40%, which indicates that biopsy accuracy needs to be improved.
Endoscopic assessment of microvessels in EGC by using magnification endoscopy combined with NBI has been studied. A quantitative assessment of the microvessels in gastric mucosa was conducted by using magnification combined with NBI. The mean calibers of microvessels in both differentiated and undifferentiated GC were significantly greater than that in the surrounding mucosa. The mean caliber of microvessels in differentiated GC was greater than that in undifferentiated GC but without significant difference. However, the high failure rate of good-quality image acquisition made almost half of the patients unsuitable for analysis (61/132). Except for assessment of microvascular caliber, other investigators tried to distinguish different GC phenotypes according to the microvascular patterns. Nakayoshi and colleagues classified the microvascular patterns of GC into 3 groups by using magnification combined with NBI: group A, fine network; group B, corkscrew; and group C, unclassified. Depressed differentiated GC was more likely to exhibit a fine network pattern, whereas undifferentiated GC was more likely to exhibit a corkscrew pattern. However, there were still 41% of the cases that could not be classified. Although magnification endoscopy combined with NBI seems to be superior to conventional endoscopy, a systematic review suggested that the consequences of its application are limited.
Other virtual chromoendoscopy technologies, such as FICE system (Fujinon, Wayne, USA) and i-Scan system (Pentax, Tokyo, Japan), were also applied in the diagnosis of EGC but their efficacy is under study. Available data indicate that virtual chromoendoscopy improves the visualization of EGC.
In summary, chromoendoscopy with conventional endoscopy and virtual chromoendoscopy are helpful in the detection of EGC. Magnification endoscopy combined with NBI not only increases the detection rate of EGC but also distinguishes different histologic phenotypes according to microvascular architecture. However, analysis of microvessels by NBI could not be performed without first obtaining a clear image of the blood vessels in some patients. Gastric pit pattern assessment by using enhanced magnification endoscopy seems to be both sensitive and specific but has poor positive predictive value. Histology remains the gold standard for diagnosis of GC.
Endomicroscopy of GC
Identification of GC by Endomicroscopy
At present, the in vivo CLE features of GC are based on fluorescein sodium staining. In vivo characterization of GC mainly consists of 3 aspects: cell, tissue structure, and microvessel. Although all the studies on in vivo diagnosis of GC by using CLE revealed excellent sensitivity from 81.8% to 92.6% and specificity from 99.4% to 100%, the diagnostic systems applied in these studies are not unified. The descriptions of in vivo CLE imaging of GC by Kakeji and colleagues mainly involve irregular variable-sized cancer cells and irregular bizarre-looking vessels. Kitabatake and colleagues paid much attention to the deformation of tissue structure, such as disorganized configuration of glands and variable shapes of ductal structures. According to Zhang and colleagues’ classification of gastric pit patterns by CLE, type G having atypical cells or glands without normal pits. For the diagnosis of GC, type G had a sensitivity of 90%, a specificity of 99.4%, an accuracy of 97.1%, a positive predictive value of 85.7%, and a negative predictive value of 97.6%. Liu and colleagues’ report focused on the altered microvessels. Unlike the regular honeycomblike microvessels in the gastric body and the coil-shaped microvessels in the antrum, the microvessels in GC are irregular in shape, number, and distribution. The differences among the diagnostic systems may be because of differences in observers’ preferences and sample selection. Systematic integrated diagnostic criteria should be recommended, which is summarized in Table 1 .