Narrow Band Imaging for Gastric Neoplasia

1 BACKGROUND

Precursor Lesions

Gastric cancer is the fourth most common cancer and second leading cause of malignant death in the world. It often presents only with vague symptoms of dyspepsia and consequently is frequently diagnosed at advanced stages.

Gastric cancer develop in a series of steps beginning with H. pylori infection. 2-4 It induces an inflammatory reaction with the surrounding gut epithelium which is theorized to drive a subsequent progression in some patients to mucosal atrophy, intestinal metaplasia, dysplasia, and finally gastric adenonocarcinoma.

Systematic biopsy protocols

In the evaluation for precancerous gastric lesions and Helicobacter pylori, experts recommend that biopsies be obtained from the antrum (3cm from pylorus), body (8cm from the pylorus), and insisura/angle. Both the greater and lesser curvature should be sampled. Additionally, recently developed staging systems including OLGIM (operative clinic on intestinal metaplasia) scores require histologic assessment (via the updated Sydney score) from two sites (antrum and corpus).

Narrow Band Imaging in endoscopy

Patients with H. pylori gastritis, gastric atrophy, intestinal metaplasia most commonly have no visible lesions on white-light endoscopy, although endoscopic findings may include antral nodularity, absent rugae, prominent gastric vessels white mucosal deposits. However, the sensitivity and specificity of these gross findings for underlying histological findings is poor. Therefore a number of image-enhancement techniques including chromoendoscopy using mucosal dyes or endoscopy-based virtual chromo-endoscopy (e.g., narrow band imaging) have been proposed. Narrow band imaging is the most widely investigated.

Narrow band imaging is an electronic, noninvasive technique in which the illuminating light from the endoscope is filtered to enable passage primarily of two narrow bands of light, 415nm and 540nm. These wavelengths correspond to the hemoglobin absorption wavelength in the capillaries and submucosal vessels respectively. This enhances evaluation of the mucosal surface patterns and vascular irregularities. NBI has been shown to be useful in the detection of dysplasia in Barrett's esophagus and characterization of small colonic adenomas.

Recently, a simple NBI classification using high-definition white light endoscopy was proposed for gastric mucosal examination.18 The NBI interpretation using this classification was compared with histological examination of mucosal biopsies, with both NBI and histology determined in blinded fashion. This classification which defines the mucosal pattern of the stomach had an accuracy exceeding 80% and excellent interobserver agreement (kappa=0.75) for normal mucosa, intestinal metaplasia, and dysplasia. However, the study was done at a referral center where 34% of patients had dysplasia and NBI was not compared with a white light assessment or standardized gastric biopsy protocol. Additionally, the results were not provided on a per patient basis, which is the most relevant endpoint in clinical practice.

2.0 OBJECTIVES AND PURPOSE Prompt detection of gastric cancer precursors enables early detection and less invasive treatment options such as endoscopic resection. Narrow band imaging is a completely noninvasive technique which uses filtered light to enhance assessment of mucosa. Our aim is to gauge whether biopsies targeted by narrow band imaging improves the detection of gastric intestinal metaplasia and gastric dysplasia relative to standard white light techniques on a per patient basis. A secondary aim will be to assess whether the technique is amenable to standardization so that it might be used more broadly to identify patients with early gastric neoplasia. While NBI is built into the vast majority of endoscopes in use few physicians are aware of its potential use.

3.0 STUDY DESIGN

The study will be a prospective tandem endoscopy trial. All EGDs will have already been planned as part of standard clinical care.

High definition white light endoscopy will initially be performed. The specific location of all mucosal findings in the stomach such as ulceration or nodularity which require biopsy will be noted by the endoscopist and research coordinator but will not be biopsied until after NBI. This is done so that blood will not distort or bias NBI assessment. Any abnormal findings in other parts of the GI tract examined using the scope (esophagus and duodenum) will be noted and biopsied.

At the completion of the white light exam, while the scope is in the stomach, the white light endoscopist will press a button on the scope which changes the view to the narrow band imaging.

At this point the NBI endoscopist who is initially blinded to the white light findings will enter the procedure room and examine the stomach using NBI. The type and location of NBI abnormalities will be noted and biopsies obtained.

At the end of the NBI exam the NBI endoscopist will switch the scope view back to white light mode. The white light endoscopist will return to the room and biopsy any sites identified and recorded during the initial white-light endoscopy. A research coordinator present for the entire procedure will verify and record that all sites identified during the initial white light exam are biopsied.

Subsequently, random biopsies will be performed by taking 2 biopsies from the lesser curvature (body and antrum), 2 biopsies from the greater curvature (body and antrum), and one from the angle.

The biopsies obtained by white light exam, narrow band imaging exam, and random sampling will be separately coded and submitted to pathology. Histologic analysis will be performed by expert GI pathologists blinded to the acquisition approach.

Short 10 second video clips of each site targeted for biopsy by white light narrow band imaging will be recorded. They will be matched with the final biopsy results and stored WITHOUT personal health identifiers. These short videos may be used for training and shared with collaborators to assess inter-observer variability and standardize the interpretation of NBI of the stomach.

The primary outcome measure will be yield of NBI, high definition white light endoscopy, and random biopsy for the detection of atrophic gastritis, IM and dysplasia on a per patient basis. A secondary endpoint will be the number of regions found by each method to exhibit atrophic gastritis, IM, and dysplasia (per lesion (region) yield). The yields of H. Pylori by method and the total number of biopsies guided per method will be additional outcomes.

Patients will be enrolled at the Los Angeles County Hospital of the University of Southern California as well as the Gastroenterology Unit at the University of Porto in Porto Portugal. The protocol originates from and statistical analysis will be done at the University of Southern California. No personal health identifiers will be exchanged at any point between the two institutions.

Prior to the formal initiation of the study there will be a lead in period of 10-20 patients with gastric symptoms. The initial patients will be examined using the white light, NBI, and gastric biopsy protocol. After this the images will be discussed by the investigators at the two centers to make certain that NBI interpretation of gastric premalignant changes is standardized. The LAC+USC investigators will also review the Portugese video training library on gastric NBI. Any changes in performance of the lead in versus the study will be noted to address the secondary aim of developing a standardized approach to NBI which may help this technique be used widely to identify patients with early gastric neoplasia.

4 STATISTICAL CONSIDERATIONS

The Fisher's exact chi squared test will be used for dichotomous outcomes such as the accurate detection of the highest level histology and number of biopsies. Adverse reactions will be reported in a descriptive manner.

Based on previous research which showed 74% correction detection of gastric cancer precursors with white light endoscopy versus 89% with NBI and given our anticipated gastric cancer prevalence of 20% we performed preliminary sample size estimates for a range of OR using G*Power (alpha=0.05, beta =0.20). We anticipate an N of 200 will be sufficient to show a significant difference between methods.