Validation of the NICE Classification Using Pentax Chromoendoscopy

Overview

Colorectal cancer (CRC) is the most frequent gastrointestinal tumor and the second cause of cancer related death. Colonoscopy is currently the recommended method for detection of polyps and cancers in the colon. Removal of all adenomatous polyps during colonoscopy has become worldwide a standard procedure as it has been demonstrated to significantly reduce colorectal cancer incidence and mortality. It is routine practice to remove all the detected polyps for pathological evaluation, due to the low accuracy (59% to 84%) to differentiate non-neoplastic from neoplastic colorectal lesions with white-light endoscopy. The development of electronic or virtual chromoendoscopy (CE) has aimed to reliably predict histology of colorectal lesions based on endoscopic features. This technology differentiates between neoplastic and non-neoplastic lesions base on the analysis of the neo-angiogenesis and the mucosal pit pattern. Optical endoscopic diagnosis allows the real-time evaluation of polyp histology during colonoscopy and to determine the appropriate therapeutic strategy. This is important in clinical practice, since adenomas or superficial invasive submucosal carcinoma lesions can be curatively treated by endoscopic removal, unlike deeply invasive carcinomas, which requires surgery. The Narrow-band imaging (NBI) international colorectal endoscopic (NICE) classification is validated classification system proposed as a valid tool for not only differentiating hyperplastic from adenomatous polyps, but also predicting submucosal deep (SM-d) carcinomas. It was developed based on NBI technology, leaving uncertainty on its applicability to other systems. It was previously evaluated the application of the NICE classification to Fujinon spectral Imaging Color Enhancement (FICE) technology founding suboptimal results (accuracy 77%, sensitivity 77% and specificity 75%) and moderate inter-observer agreement (kappa: 0.51).

NBI technology uses a physical filter in order to exploit the capacity of hemoglobin to selectively absorb blue light. It is considered to be less useful in areas with a large lumen, such as the stomach and colon, as a result of insufficient light for wide-range observation of the full extent of the tissue surface, and because the images are dark. I-scan is a virtual chromoendoscopy, based on the principle of digital post-processing and consists of three different image algorithms. Mode 1 is for detection of lesions. This algorithm is used for surface enhancement (SE) and enhances light-dark contrast by obtaining luminance intensity data for each pixel and apply an algorithm that allows detailed observation of the mucosal surface structure and lesion borders without altering the brightness of the endoscopic picture. Mode 2 is for characterization of lesions. This algorithm combines SE and tone enhancement (TE). TE dissects and analyses the individual red-green-blue components of a normal endoscopic image in real time and then alters the color frequencies of each component and recombines the components to a single, new color image without visible delay for the examiner. It is used to enhance minute mucosal changes and vessel structures. Mode 3 adds contrast enhancement (CE) to the endoscopic image (in addition to SE and TE) and is for demarcation of lesions. It digitally adds blue color to dark areas within the endoscopic image. Recently an image-enhanced endoscopic technology using a pre-processor band-limited light called Optical Enhancement system (OE system™), was developed by HOYA Co. (Tokyo, Japan) and is now equipped with the latest endoscopy system (Pentax Video Processor EPK-i7010; HOYA Co.). This new technology combines digital signal processing with optical filters that overcomes the limit of the spectral characteristics of the illumination light. Previous i-scan technology uses white light alone as an illumination light and digital post-processing of the reflection afterwards creates images yielding the virtual chromoendoscopic image. Although accumulating evidence has shown the usefulness of i-scan in the clinical setting, emission of white light alone causes a potential limitation for the current i-scan technology to obtain higher contrast images of microvascular pattern on the mucosal surface as shown by Narrow Band Imaging (NBI). The basic concept of OE is to overcome the darkness of NBI, which results in less usefulness for detectability in wide-range observation in the full-extended gastrointestinal lumen. The new innovated optical filters achieve higher illumination intensity and overall transmittance by connecting the peaks of the hemoglobin absorption spectrum (415 nm, 540 nm and 570 nm) creating a continuous wavelength spectrum. There are two modes with different OE filters. Mode 1 is designed mainly to improve visualization of microvessels with a sufficient amount of light, and Mode 2 is designed to improve contrast of white-light observation by bringing the color tone of the overall image closer to that of natural color (white color tone). Due to the underlying differences between the NBI and Pentax technologies (i-scan and OE system), it remains uncertain whether the NICE classification may be translated to this technology. The aim of this study is to validate the NICE classification by applying i-scan and OE system with high-definition without optical magnification, to evaluate colorectal lesions. MATERIALS AND METHODS Study design: It will be a non-interventional, prospective, non-randomized, non-controlled and simple blind study, performed in the Ecuadorian Institute of Digestive Disease (IECED), OmniHospital Academic Tertiary Center Ecuador, with patients included from April 2017 to October 2017. The study protocol and consent form has been approved by the institutional review board and will be conducted according to the declaration of Helsinki. Written informed consent will be obtained from all subjects before the examination. Study Population: Consecutive adult patients between 18 and 80 years of age, referred for elective outpatient colonoscopy and in whom polypectomy or biopsy is performed will be enrolled. Exclusion criteria will be pregnancy, suspected colonic obstruction or history of previous obstruction, gastrointestinal bleeding, history colorectal surgery, inflammatory bowel disease, hereditary polyposis syndrome, diverticulitis, history of radiation therapy to abdomen or pelvis, history of severe cardiovascular, pulmonary, liver or renal disease, severe coagulation disorders or use of anticoagulants. Patients with polyps but in whom histopathology has not been evaluated or with a poor bowel preparation (Boston Bowel Preparation Scale ≤6) will be excluded from the analysis but included in the intention to treat. Intervention: endoscopic technique The procedures will be performed under propofol intravenous sedation, in left later or supine position. For bowel preparation, the participants will have to ingest 4 liters of polyethylene glycol solution in the evening the day before the procedure. The effectiveness of the bowel cleansing will be evaluated using the Boston Bowel Preparation Scale. Three colonoscopists (C.R.M., M.V., M.S.A.), with extensive experience in colonoscopy with i-scan and OE (\>1000 cases) and familiar with the NICE classification, will perform the procedures using Pentax high definition colonoscopes without optical magnification (EC-3890Zi, PENTAX Medical, HOYA Co.) and EPK-i7010 processor with the ability to display i-scan and OE system™ images. The endoscopy images will be analyzed on a 27-inch, flat panel, high definition LCD monitors (Radiance™ ultra SC-WU27-G1520 model). All lesions initially detected by white light endoscopy will be intensively washed using a water ejection pump before recording the procedure and then the endoscopic features on the surface will be evaluated using the 3 i-scan modes and the 2 OE modes without magnification. The lesion size data (1-5/6-9/\>10 mm), location (cecum/ right/ transverse/ left/ sigmoid colon/ rectum) and macroscopic shape of the lesions based on the Paris classification will be recorded. The size will be estimated with biopsy forceps (2.2 mm closed; Radial Jaw 4, Boston Scientific, Marlborough, Massachusetts, USA) or polypectomy snare (13mm open; Captivator, Boston Scientific, Marlborough, Massachusetts, USA). Finally all lesions will be classified in real-time into 3 types based on NICE classification (NICE 1, hyperplastic polyps; NICE 2, adenoma and superficial submucosal carcinoma; NICE 3, SM-d invasive carcinoma). A level of confidence (high or low) will be assign in each stage. A high confidence prediction will be considered when the endoscopist is 90% certain of the diagnosis and this condition will be consider when polyps have ≥1 features associated with one NICE type and no features associated to the others NICE type. If there are uncertainty regarding the features or if there are features from different NICE types the prediction will have low confidence. Polyp's images will be photographically and videotape recorded. Each high-definition video will consist of 30 to 60 seconds of white-light endoscopy followed by i-scan 1,2,3 and OE mode 1 and 2. For the purpose of this study, magnification was not allowed during recording. All polyps will be resected or biopsied for histopathological examination used as the criterion standard for the analysis. Two experienced pathologists, blind to the endoscopic diagnosis, will assess the histology according to the revised Vienna classification. Lesions identified histopathologically as serrated adenomas/ polyps or traditional serrated adenoma will be excluded from the analysis but included in the intention to treat, due to the lack of fully established evidence of NICE classification utility for the diagnosis of these lesions. In a second phase, endoscopists with less experience (\<1000 procedures) using i-scan and OE will be selected. After a formal instruction on the NICE classification based on theoretical background and a series of i-scan/OE-polyp images, each of them will have to reviewed all the videos and applied the NICE classification in order to predict the histology. Each of the three criteria of the NICE system (colour/vessel/surface pattern) will be individually scored as well as the overall level of confidence (high/low). Statistical analysis Base line characteristics will be expressed as percentage or mean +/- standard deviation. It will be calculated the accuracy, sensitivity, specificity, negative and positive predictive values with the 95% of Confidence Interval (95% CI), for each component of the classification and for the overall prediction by using the classification. Using multilevel logistic regression, the sensitivity and specificity of the different criteria, will be compared. Diagnostic values of the criteria used in combination (combination of "at least 1 criterion being positive" versus "all combined criteria being positive") will be assessed and compared. The criterion standard for validation of predictions will be the lesions histology. Presence of adenomatous feature at each criterion will be defined as a positive result. The sample size was calculated assuming that 80% of predictions will be made with high confidence, and that the real accuracy will be 90%. A data set containing 30 random-selected videos will be presented after 2 months to the three main investigators (C.R.M, M.V, M.S.A.) in order to assess intra and inter-observer reproducibility. The endoscopists will have classified again the polyps according to the three types on the NICE classification. To examine inter and intra observer agreement, kappa values will be calculated. Kappa coefficients below 0.4 indicate "poor agreement," values between 0.4 and 0.8 represent "moderate to good agreement," and values greater than 0.8 indicate "excellent agreement." A P value of less than 0.05 will be considered to be statistically significant. All the statistical analysis will be performed using SPSS software suite v.22.