Confocal Laser Endomicroscopy For Early Gastric Cancer And Gastric Intraepithelial Neoplasia

Background and aimsAlthough its incidence is gradually decreasing,gastric cancer(GC) is still one of the most common cancers worldwide.It is well known that the prognosis of GC depends heavily on the stage at which the diagnosis is made.Five-year survival rate of advanced gastric cancer is very low(about 20%-30%),but early gastric cancer(EGC) may be curable with reported five-year survival rates of over 90%.Early gastric cancer is defined as an adenocarcinoma confined to the mucosa or submucosa regardless of lymph node metastasis.In Japan,up to 60%of gastric cancers are diagnosed as early cancers.However,EGC is much less frequently detected in our country,only accounting for less than 10%of all gastric cancers.Gastric intraepithelial neoplasia(GIN) is considered the most advanced preinvasive stage in the multistep morphogenesis of gastric cancer.GIN represents a direct precursor of cancer,and also a marker of risk for cancer elsewhere in the gastric mucosa.Identification and surveillance of GIN could lead to early detection of GC.It is evident that endoscopy is the best method to achieve the diagnosis of GIN and EGC,which enables the detection of subtle mucosal abnormalities and simultaneous biopsy for histological assessment.Because of nonspecific endoscopic appearance, the identification of GIN and EGC is still a great challenge in clinical practice. Although advanced endoscopic techniques may improve the detection of GIN and EGC,additional biopsy is usually needed for creating the final diagnosis.In other words,endoscopy plus biopsy remains the gold standard.Obviously,an in vivo histological diagnosis during endoscopy would be a great aid in detecting GIN and EGC.Moreover,a heavy burden of work for pathologists will not be forced to bear, and local fibrosis after biopsy will be avoided,which can make endoscopic resection difficult since adequate lifting of the lesion by submucosal injection is precluded.Confocal laser endomicroscopy(CLE) is a newly developed imaging technology. It is the integration of a miniaturized laser scanning confocal microscope into the distal tip of a conventional endoscope.CLE can produce 1000-fold magnification virtual histological images of the gastrointestinal surface and subsurface tissue,with no need for taking biopsy.However,to date,there have been no studies on the diagnostic value of CLE for GIN and EGC.The aim of the present study is to study the CLE characteristics of various kinds of gastric mucosal diseases,to establish the diagnostic criteria of CLE for gastric mucosal diseases,and to assess the true diagnostic value of CLE for gastric mucosal diseases,in particular for GIN and EGC.Methods1.In PartⅠ,the patients who underwent CLE procedure between June 2006 and November 2006 were collected,and those who had histological findings from targeted biopsy specimens were selected into the study.The CLE images of gastric mucosal lesions were reviewed and compared with the corresponding histological sections in an unblinded manner by the experienced endoscopists and pathologists.Consequently, the CLE imaging criteria were developed to grade superficial gastric mucosal lesions.2.In PartⅡ,patients with previously diagnosed polyps were recruited between January 2007 and August 2008.Inclusion criteria:gastric semi-pedunculated or sessile polyps of at least 0.5cm in their greatest diameter with hyperplastic or adenomatous characteristics conformed by endoscopic forceps biopsies.Exclusion criteria:gastric polyps of sizes＜0.5cm or＞3cm in diameter,or associated with known carcinoma,or patients with liver cirrhosis,impaired renal function,acute gastrointestinal bleeding,coagulopathy,esophageal varices,jaundice,age＜18 years old,pregnancy,breastfeeding,inability to provide informed consent,and known allergy to fluorescein sodium.Written informed consent was obtained from all participating patients before the procedure.The CLE was conducted by three experienced endoscopists(YQL,XMG and TY),who didn't know the histological diagnosis of the examined gastric polyps.In vivo CLE diagnosis was made immediately at the time of the procedure by the endoscopist according to the established CLE imaging criteria for gastric hyperplastic polyps or adenomas. Afterwards,endoscopic resection was performed to remove gastric polyps.Resected specimens were retrieved and sent for histological examination.All the collected CLE images were analyzed in combination with the histological findings to study the CLE characteristics of gastric hyperplastic polyps or adenomas.Then,the CLE images corresponding to each lesion was numbered and randomized.Three experienced endoscopists blinded to the related information about the lesions were asked to make their CLE diagnoses independently.The results of CLE diagnoses were compared with the histological findings,and the accuracy of CLE diagnosis was calculated.3.In PartⅢ,from January 2007 to August 2008,consecutive patients underwent CLE examinations.Patients scheduled for gastroscopy were enrolled for CLE owing to the following reasons:with definite or suspicious early gastric cancer lesions,with gastric precancerous conditions,or with the upper gastrointestinal(GI) symptoms. Exclusion criteria:patients with lesions diagnosed as obviously advanced GI cancer, patients with acute gastrointestinal bleeding,liver cirrhosis,impaired renal function, coagulopathy,ascites,esophageal varices,jaundice,80 years old＜age＜18 years old, pregnancy,breastfeeding,inability to provide informed consent,and known allergy to fluorescein sodium.Written informed consent was obtained from all participating patients before the procedure.All the endoscopic and histological findings were recorded and stored in a database.Endoscopic findings were described according to the updated Sydney System of gastritis,the Paris endoscopic classification of superficial neoplastic lesions in the stomach and the Borrmann classification of gastric cancer.CLE images were stored in computer folders corresponding to each patient on compact disks.High-quality WLE pictures were captured and stored by the Medical Imaging System.Targeted biopsies were taken for the abnormal lesions showed by CLE.Patients who had point-to-point targeted biopsy were selected for analysis. Three experienced endoscopists blinded to the histological findings were asked to make a CLE diagnosis independently according to the CLE imaging criteria for gastric mucosal lesions.Compared with the histological diagnoses,the accuracy of CLE diagnosis was calculated.Intra-and inter-observer agreement was also accessed.4.In PartⅣ,consecutive patients of aged over 45 years old were recruited into the prospective study between September 2008 and September 2009.Inclusion criteria: 45 years old≤age＜80 years old;patients with dyspeptic symptoms or with gastric pre-malignant conditions for surveillance gastroscopy.Exclusion criteria:patients with known cancers or gastrectomy;patients scheduled for endoscopic treatment; patients with alarm symptoms such as serious anaemia,dysphagia,and marked weight loss;and patients under conditions unsuitable for performing CLE such as fever, ascites,jaundice,varices,coagulopathy,liver cirrhosis,impaired renal function,acute gastrointestinal bleeding or obstruction,pregnancy,breastfeeding,known allergy to fluorescein sodium,inability to provide informed consent and other situations.Written informed consent was obtained from all the enrolled patients before the procedure. Patients scheduled for gastroscopy were interviewed,and those in line with the inclusion and exclusion criteria were selected.The selected patients were numbered, and some of which were sampled and assigned to group A and B at 1:1 ratio.In group A,conventional gastroscopy was performed;and in group B,CLE was performed. Two experienced endoscopists conducted endoscopy and alternated in group A and B, making in vivo CLE diagnosis according to the CLE imaging criteria.Targeted biopsy was performed when lesions with IM,GIN or cancer was considered.Results1.In PartⅠ,data for 198 patients with gastritis(68),IM(54),ulcer(8),low grade intraepithelial neoplasia(LGIN,19),high grade intraepithelial neoplasia(HGIN,6), adenocarcinoma(26) and 17 healthy volunteers were selected.Various kinds of gastric mucosal diseases had their corresponding characteristics in CLE images.In CLE images from the antrum,cobble-like appearance of normal mucosa pattern was seen;continuous short rod-like pits with slit-like openings of foveolae and coil-shaped subepithelial capillary networks could be easily identified.With regard to the body and fundus with oxyntic glands,round pits with round openings of foveolae and honeycomb-like submucosal microvasculatures were clearly observed.Inflammation was identified on the basis of infiltrating inflammatory cells,raising vessels,oozing fluorescein sodium,or elongated and tortuous pits;atrophy was identified based on the number of pits decreasing and pits prominently dilating;IM was identified when "goblet cell"(large black cell contrast to surrounding columnar-lined epithelium cells), or "villiform shape",or "absorptive cell"(more slender,and brighter) were observed. The CLE images of gastric cancer were obvious different from the above lesions,and showed that the architecture of normal gastric pits disappeared and was replaced by obviously abnormal black cells and disorganized architecture representing carcinoma.2.In PartⅡ,a total of 44 patients with 60 gastric hyperplastic polyps and 21 patients with 27 gastric adenomas were eligible for analysis.Compared with the surrounding background mucosa,most of gastric hyperplastic poly(?)s(49/60,81%) showed marked hyperplasia of surface columnar epithelium cells covering gastric pits nearest the lumen,which were identical to the normal columnar epithelium cells without any atypia by CLE.Hyperplastic columnar epithelium cells regularly arranged and encircled the openings of elongated or branch-like pits.Varied forms with foveolar sulci in different dilated degree and length could be seen in confocal images.In a single field of view(FOV),the number of gastric pits decreased obviously as a result of epithelium cell hyperplasia.In addition,interstitium was broadened and partly brightened as a result of the excess of edematous mucosa; moreover,interstitium was often infiltrated by inflammatory cells.However,a small amount of hyperplastic polyps(7/60,12%) showed a finger-like surface pattern;and the surface pattern of the other four(4/60,7%) indicated moderate epithelium cell hyperplasia,and mildly dilated or elongated foveolae.Occasionally,IM or atypical cells could be seen in a few confocal images.Gastric adenomas could be clearly imaged by both fluorescein aided CLE and acriflavine aided CLE.During the fluorescein aided CLE imaging,a distinct cell type with a tessellated appearance known as "atypical cell" usually composing IN,showed itself as a type of black cell with irregular shape encircled with white interstices,significantly different from regular columnar epithelium cell.During the acriflavine aided CLE imaging,atypical cells were present as high grayscale cells with irregular shape,nonuniform size,and enlarged nuclei.Moreover,gastric adenoma,whether tubular or villous adenoma, showed significant architectural changes in confocal images.Irregular ridges or villi, which sometimes looked like "cerebriform" shape,were usually observed.Focal asymmetrical ridge distortion was also often seen,implying the presence of dysplasia. Distorted or ridgelike openings of glands in different length and width were often observed when tubular components predominated.Broadened and brightened interstitium was seldom observed.In addition,another distinct cell shape with a light gray appearance similar to "absorptive cell" composing intestinal metaplasia was also observed in confocal images,which was more slender than normal gastric columnar epithelium cell.The in vivo CLE diagnosis was made as hyperplastic polyps or adenomas.The overall accuracy of CLE diagnosis was 90%and the kappa coefficient of agreement between histopathology and in vivo CLE imaging was 0.77.Intra-and inter-observer agreement was 0.92 and 0.83,respectively.If the final CLE diagnosis was determined when two endoscopists were of the same opinion about the same lesion,the overall accuracy was 97%.Gastric adenomas could be further divided into LGIN and HGIN by CLE.All the adenomas with LGIN were classified into LGIN group,but 2 of 4 adenomas with HGIN were attributed to LGIN group mistakenly.3.In PartⅢ,data for 182 patients were eligible for analysis.In total,42 patients had EGC,9 HGIN,30 LGIN,52 gastritis,and 49 IM.Patients with EGC or GIN had single lesions,and each patient with gastritis or IM was considered to have a single lesion.37 EGC cases were determined by histology findings after surgical ablation;5 EGC and 9 HGIN cases were determined after endoscopic resection.Of 42 EGCs,25 were DCA(14 well-differentiated and 11 moderately differentiated adenocarcinoma), and 17 were UCA(12 poorly differentiated adenocarcinoma and 5 signet-ring cell carcinoma);13 were intramucosal carcinomas and 29 were carcinomas with submucosal invasion.EGC could be determined by CLE with high sensitivity(88.1%) and specificity(98.6%),and gastric neoplastic lesions(EGC and GIN) could be distinguished from non-neoplastic lesions by CLE with 84.0%sensitivity and 92.1% specificity.However,GIN(HGIN and LGIN) was recognized by CLE with low sensitivity(66.7%) and high specificity(92.3%).The distinction between LGIN and gastritis was unsatisfactory,and HGIN was easily confused with carcinoma by CLE. In addition,well-differentiated adenocarcinomas were sometimes misdiagnosed as IM by CLE.Although the characteristics of CLE images vary by diseases-normal mucosa,gastritis,IM,IN or carcinoma,gastric mucosal lesions could be classified into 2 categories:"cancerous" lesions of HGIN or adenocarcinoma and non-cancerous lesions of normal mucosa,gastritis,IM,or LGIN.According to the two-tiered system, "cancerous" lesions could be distinguished from non-cancerous lesions by CLE with high sensitivity(90.2%),specificity(98.5%),PPV(95.8%),NPV(96.3%),and accuracy(96.2%).Identification of cancerous lesions was based on irregular and abnormal signs of architecture,cells,or microvessels on CLE.Abnormal architecture was characterized by loss of regular surface patterns and the appearance of atypical glands or disorganized patterns;atypical cells were often dark,irregular in shape and size,and disordered;and abnormal microvessels were rigid or irregular,with increased caliber and unusual shape.Abnormal cytology features combined with changes in structure were often used to identify cancerous lesions.On the basis of CLE images alone,DCA and UCA could be determined with an accuracy of 84.9%. The diagnostic validity could be improved with the addition of WLE images(accuracy, 94.4%).But,no statistical differences were observed(P＞0.05).Intra-and inter-observer agreement was 0.827 and 0.783 for differentiating "cancerous" and non-cancerous lesions,0.870 and 0.838 for differentiating DCA and UCA by CLE.4.In PartⅣ,a total of 2130 patients coincided with inclusion criteria and failed to meet exclusion criteria were enrolled in the prospective study.There were 1065 patient in group A and 1065 patients in group B.There were no significant differences observed between group A and B regarding sex,age and gastric pre-malignant conditions.Patients with pre-dominant duodenal or esophageal lesions in need of biopsies or with esophageal or fundic varices were excluded from the study.In the end, 1004 patients in group A(8 patients with duodenal lesions,47 patients with esophageal lesions and 6 patients with varices were excluded) and 1003 patients in group B(6 patients with duodenal lesions,53 patients with esophageal lesions and 3 patients with varices were excluded) were used for analysis.CLE targeted biopsies were performed for 524 lesions in 414 patients(27 patients with advanced GCs were not included for lack of CLE examination).Based on the two-tiered system,cancerous lesions could be determined by real-time in vivo CLE with a sensitivity of 99.4%,a specificity of 91.1%and an accuracy of 98.5%.Gastric adenocarcinoma could be further classified into DCA and UCA by real-time in vivo CLE.Among 54 lesions diagnosed as cancerous lesions by CLE,DCA and UCA could be determined with sensitivity of 61.5%and 100%,specificity of 92.9%and 77.8%,PPV of 88.9%and 69.2%,NPV of 72.2%and 100%,accuracy of 77.8%and 85.2%,respectively.A total of 140 malignant neoplasms were identified in group A(64) and group B(76),4 cases of which were excluded from analysis(1 lymphomas,1 malignant stromal tumor in group A;1 lymphoma and 1 carcinoid in group B).Staging of 13 lesions(group A,8; group B,5) were determined by EUS.In the end,5 EGCs(EGC/GC,5/62) were detected in group A,and 23 EGCs(EGC/GC,23/74) were found out in group B. There were 3.9 times more detected EGCs in group B than in group A(P＜0.01).No statistical differences were observed on gross morphology or pathological types of the detected EGCs between the two groups.Conventional gastroscopy guided biopsies were performed for 360 lesions in 264 patients in group A,and CLE targeted biopsies were done for 524 lesions in 414 patients in group B(27 patients with advanced GCs were not included for lack of CLE examination).Much more patients with IM or LGIN were found out in group B than in group A(IM,271/1003 vs 120/1004;LGIN, 65/1003 vs 33/1004)(P＜0.01).Much more IM and LGIN lesions(IM,341 vs 141; LGIN,98 vs 37) were detected in group B than in group A(P＜0.01).As far as visible lesions were concerned,targeted biopsies were performed for 960 lesions(group A vs group B;360/1004 vs 300/1003;P=0.005) in 502 patients(group A vs group B; 264/1004 vs 238/1003;P=0.184).There were more non-metaplastic and non-neoplastic lesions detected in group A than in group B.More lesions with GIN (LGIN and HGIN) were detected in group B than in group A(P＜0.01),in spite that there were no differences relating to the detected lesions with IM or HGIN(P＞0.05).Conclusions1.Various kinds of gastric mucosal diseases,such as normal mucosa,gastritis,IM, IN or cancer,had their characteristics in CLE images.The CLE imaging criteria for gastric mucosal lesions were developed.CLE can predict gastric mucosal histology.2.Gastric hyperplastic polyps and adenomas had different characteristics in CLE images;and could be identified by CLE during or after endoscopy with high accuracy.3.Gastric cancer could be determined and classified into DCA and UCA with high validity and reliability by real-time in vivo CLE or following CLE after endoscopy.4.A two-tiered CLE classification for gastric mucosal lesions was proposed. Lesions both suspected of being carcinoma and predicted as HGIN or carcinoma by CLE were regarded as "cancerous" lesions,and non-cancerous lesions defined by CLE included normal mucosa,gastritis,IM,and LGIN.Identification of cancerous lesions was based on irregular and abnormal signs of architecture,cells,or microvessels in CLE images.Abnormal architecture was characterized by loss of regular surface patterns and the appearance of atypical glands or disorganized patterns; atypical cells were often dark,irregular in shape and size,and disordered;and abnormal microvessels were rigid or irregular,with increased caliber and unusual shape.Abnormal cytology features combined with structure changes were often used.5.More patients with GIN and EGC could be detected by CLE than by conventional gastroscopy.No statistical differences were observed on gross morphology or pathological types of the detected EGCs between the two groups.CLE could be a useful means for screening and surveilling high risk populations of GC.SignificanceIn the present study,the CLE imaging criteria for gastric mucosal lesions were firstly established and accessed;and a prospective randomized controlled study was firstly designed and performed to determine whether CLE can improve the detection of EGC and GIN as compared to conventional gastroscopy.Anyway,CLE is such an attractive technology that the real-time',in vivo,instant,accurate histology diagnosis will give great benefit to the diagnosis and treatment of gastrointestinal diseases.