The reseachers’ objectives were to assess the accuracy of predicting histology (tissue composition or polyp type) by using narrow band imaging in real time for distal colorectal polyps. It was a prospective, observational study conducted at Indiana University Hospital and their affiliated ambulatory surgery center. The study involved 225 consecutive adults undergoing elective screening or surveillance colonoscopy. Researchers evaluated real-time histology of 235 distal colorectal polyps from 31 patients by using high-definition colonoscopy and NBI without optical magnification. For each polyp, the endoscopist described size and surface characteristics (vascular and pit pattern, color, pseudodepression). Before removal, histology was predicted, and a level of confidence (high or low) was assigned. The main outcome measurements were sensitivity and negative predictive value of high-confidence endoscopic predictions of adenomatous versus hyperplastic histology for polyps less than or equal to 5 mm in size. All colonoscopies were performed by a single experienced endoscopist by using high-definition colonoscopes.
The histology of 216 of the 235 polyps, or 91.9 percent, was predicted with high confidence. Overall, the histology of 220 of the 235 polyps was predicted correctly in real time, representing an accuracy of 93.6 percent. When the prediction was made with high confidence, the accuracy was 97.7 percent (211/216). The accuracy of high-confidence, real-time predictions for polyps less than or equal to 5 mm was 99 percent (199/201). Of 38 adenomas in the sigmoid colon and rectum, only 4 were incorrectly predicted to be hyperplastic. Only 11 of 197 hyperplastic polyps were incorrectly predicted to be adenomas. This represents an overall sensitivity for real time evaluation of 89.5 percent and specificity of 94.4 percent.
Performance characteristics improved when predictions were made with high confidence and when polyps less than or equal to 5 mm were considered. When predictions were made with high confidence, the sensitivity was 93.9 percent, and the specificity was 98.4 percent. For polyps less than or equal to 5 mm predicted with high confidence, the sensitivity was 96 percent and specificity 99.4 percent. The positive predictive value of an adenoma prediction was 75.6 percent, and the negative predictive value was 97.9 percent. Predictive values were the greatest for high confidence predictions with polyps less than or equal to 5 mm, where there was a positive predictive value of 96 percent and a negative predictive value of 99.4 percent.
The study was limited by being conducted at a single center by a single endoscopist with a special interest in colonoscopy and extensive experience in NBI. According to the researchers, this is the first prospective study and the second study overall to specifically address the ASGE Preservation and Incorporation of Valuable Endoscopic Innovations (PIVI)* on leaving distal colon hyperplastic polyps in place, and both studies have found that NBI allows the recommended ASGE threshold for negative predictive value to be exceeded by a substantial margin.
The researchers concluded that an expert using NBI and high-definition optics can achieve an accuracy of real-time assessment of distal colon diminutive polyps that substantially exceeds the accuracy threshold recommended by the ASGE for a new management paradigm of leaving diminutive hyperplastic rectosigmoid polyps in place. The researchers recommend that NBI be tested for real-time histologic assessment of distal colorectal polyps by a larger group of physicians in a variety of practice settings.
*See the ASGE Preservation and Incorporation of Valuable endoscopic Innovations (PIVI): Real-Time Endoscopic Assessment of the Histology of Diminutive Colorectal Polyps (March 2011) http://www.asge.org/clinicalpractice/clinical-practice.aspx?id=11958