Characterization of real-time narrow-band imaging of diminutive colorectal polyp histology: A Chinese national nonaffiliated multicenter investigation

To the Editor: Colorectal cancer ranks third globally, with rising Asian prevalence. Diminutive polyps typically have drastically reduced the risk of advanced adenomatous histology; however, they are regularly monitored to exercise caution. A noninvasive monitoring method, such as real-time optical assessment, instead of diminutive polyp histology, can greatly benefit patient care while reducing medical expenses without compromising clinical decision-making or quality. The “resect and discard” and “diagnose and leave” strategies require an optical diagnosis of ≥90% negative predictive value (NPV) for rectosigmoid adenomatous polyps and ≥90% surveillance interval agreement with the guidelines set by the American Society for Gastrointestinal Endoscopy (ASGE). Narrow-band imaging (NBI) enables “resect and discard” strategies, requiring ≥90% NPV for rectosigmoid polyps and ≥90% surveillance interval agreement as per the ASGE guidelines.[1] To date, several investigations have examined the possibility of trained endoscopists achieving the aforementioned goals. An extensive meta-analysis involving 56 research articles examining NBI for optical diagnosis revealed an overall sensitivity of 91.0% (95% confidence interval [CI] 88.6–93.0%), specificity of 85.6% (95% CI 81.3–89.0%), and NPV of 82.5% (95% CI 75.4–87.9%).[2] A meta-analysis of optical diagnosis of diminutive polyps revealed that the diagnostic precision was higher in the academic centers, particularly when conducted by experienced endoscopists. In contrast, only 3 of 20 NBI investigations took place in nonacademic settings,[2] and few investigations to date have evaluated the feasibility of optical diagnosis of diminutive polyps in Chinese settings. In this report, we investigated the performance of nonacademic Chinese gastroenterologists, without prior NBI training, in achieving 90% NPV for adenomatous polyps with high confidence (HC) in rectosigmoid polyp diagnosis and 90% agreement in surveillance intervals when combining HC NBI estimations or histological verification of low-confidence (LC) NBI estimations following routine NBI training in China. Twenty-four gastroenterologists from 12 participating nonaffiliated hospitals (11 tertiary grade A hospitals and 1 tertiary grade C hospital) across China. Eighteen doctors had 2–5 years of colonoscopy experience, whereas six doctors had over 5 years of colonoscopy experience. All hospitals used Olympus EVIS LUCERA processors (Olympus Corporation, Tokyo, Japan) with high-definition monitors and Workgroup serrAted polypS and Polyposis (WASP) stratification posters 21. The study was approved by institutional review boards (No. 2018-P2-088-01) and registered in Clinical Trials (NCT03903497). Gastroenterologists received a 40-minute audiovisual NBI training module covering NBI principles, the WASP stratification system,[3,4] and quality image capture. Participants were required to have achieved ≥90% accuracy in the posttraining examination within two attempts. Successful participants learned the “characterize, resect, and discard” approach and ASGE Preservation and Incorporation of Valuable Endoscopic Innovations (PIVI) thresholds. Patients over 18 years with ≥1 diminutive polyp (April 2019–May 2023) were prospectively enrolled. Exclusion criteria included poor bowel preparation (total Boston bowel preparation scale score <5 or any segment score <2), earlier colorectal surgery, familial adenomatous polyposis, inflammatory bowel disease, advanced cancer, or missing histological information. During colonoscopy, subjects conducted real-time optical diagnosis for diminutive polyps using NBI-based stratification. Polyp size, location, NBI patterns, predicted histology, and confidence degree (HC vs. LC) were documented. Polyps were removed for individual histological analysis. Surveillance interval predictions were made immediately following HC optical diagnosis. For LC profiling, predictions relied on polyp histology and HC optical diagnosis of diminutive polyps. Predictions followed the American recommendations for colonoscopy surveillance. Primary outcomes included: (1) surveillance interval agreement for patients with ≥1 diminutive polyp diagnosed with HC; (2) NPV of diminutive rectosigmoid polyps with HC identification; and (3) WASP stratification diagnostic precision. The sample size was based on the largest sample size for the three major outcomes. We speculated that achieving a 93% NPV HC diagnosis, as per WASP stratification, with an additional 3% precision, would require 920 cases of diminutive (≤5 mm) hyperplastic polyps. As 20% of polyps are diminutive rectosigmoid polyps, 1643 independent diminutive rectosigmoid polyps were needed. As these polyps represent 42% of all diminutive polyps, 3912 total diminutive colorectal polyps (CPs ≤5 mm) were required. As, on average, 1.5 diminutive polyps are detected per colonoscopy, 2608 colonoscopy patients were needed. Details of these computations are provided in the Supplementary Materials, https://links.lww.com/CM9/C832. For assessment of accuracy, sensitivity, specificity, positive predictive value (PPV), and NPV, adenoma represented a positive outcome, whereas nonadenoma represented a negative outcome. Sessile-serrated lesions (SSLs) were defined as “adenomas.” Surveillance interval predictions derived from NBI characterizations, as well as actual surveillance interval recommendations given to the patient, were compared to determine agreement in surveillance intervals. Data were analyzed using SAS (SAS Institute Inc., Cary, North Carolina, USA) and R software (v.4.2.2, R Foundation for Statistical Computing,Vienna, Austria). The study subjects conducted 2641 colonoscopies to identify diminutive polyps. As shown in Supplementary Table 1, https://links.lww.com/CM9/C832, the surveillance interval was estimated immediately after the procedure in 1516 cases (57.4%). In an additional 1125 cases (42.6%), histology was necessary for LC profiling or for polyps >5 mm. These procedures were evaluated by a researcher, with estimations according to the NBI predictions and accompanying histology. As depicted in Supplementary Table 2, https://links.lww.com/CM9/C832, of the 3129 diminutive polyps examined, 2401 (76.7%) were profiled with HC, and 1974 (63.1%) were designated as tubular adenomas. Moreover, 45 (1.5%) were diagnosed as SSLs, with no dysplasia, whereas 19 (0.6%) diminutive polyps were high-grade dysplasia or cancer. Additionally, 1403 polyps were found in the rectum or sigmoid colon, of which 1108 (79.0%) were profiled with HC. Study subjects achieved an NPV of 93.2% (95% CI 91.7–94.7%) in the rectosigmoid colon for polyps with HC. The HC surveillance interval prediction using NBI-based diminutive polyp optical diagnosis, along with histology for LC profiling and polyps >5 mm, was 91.7% (95% CI 90.7–92.8%). As shown in Figure 1A, among the 218 procedures with surveillance interval disagreement, 78.9% (n = 172) estimated a longer interval, and 21.1% (n = 46) estimated a shorter interval, compared to the histological results; therefore, 93.5% (2469/2641) of estimations would encourage patients to seek timely or early follow-up or monitoring. Additionally, the WASP stratification accuracy in optical adenoma diagnosis was 87.6% (95% CI 86.3–88.9%).Figure 1: (A) Matrix of optical predicted and histological determination of surveillance intervals. (B) Each medical center’s performance on PIVI thresholds: NPV for high-confidence diagnoses in the rectosigmoid and agreement in predicted and actual surveillance intervals (% with 95% CI). CI: Confidence interval; NPV: Negative predictive value; PIVI: Preservation and Incorporation of Valuable Endoscopic Innovations.Individually, 8 out of 12 endoscopic locations (67%) achieved 90% NPV in the rectosigmoid region. Among the remaining four locations, all achieved an agreement above the 90% threshold [Figure 1B]. In addition, 9 out of 12 patients (75%) achieved an overall agreement in surveillance interval of 90%. Among the remaining three, all endoscopy locations achieved agreement percentages that were above the 90% threshold with a 95% CI. Supplementary Table 3, https://links.lww.com/CM9/C832, summarizes the NBI optical diagnostic performance for diminutive polyps by confidence level and location. The overall accuracy, sensitivity, and NPV for diminutive polyps profiled using NBI were 78.6% (95% CI 77.2–80.0%), 86.9% (95% CI 85.2–88.6%), and 71.7% (95% CI 70.1–73.2%), respectively. In case of polyps profiled with HC, subjects achieved an overall precision of 87.6% (95% CI 86.3–88.9%), sensitivity of 94.6% (95% CI 92.9–96.3%), and an NPV of 88.3% (95% CI 87.0–89.6%) for adenomatous polyps. We demonstrated markedly enhanced performance either by location or overall for CPs profiled with HC vs. LC. Supplementary Table 4, https://links.lww.com/CM9/C832, presents the WASP stratification in the optical diagnosis of SSLs. The overall accuracy and sensitivity for diminutive polyps profiled using NBI were 99.2% (95% CI 98.9–99.5%) and 79.2% (95% CI 78.9–79.4%), respectively. In case of HC-profiled polyps, participants achieved an overall accuracy of 99.6% (95% CI 99.4–99.9%), and a sensitivity of 75.0% (95% CI 74.8–75.2%) for SSLs. Our extensive multicenter prospective nonacademic investigation examining NBI usage in optical diagnosis by NBI-trained only endoscopists revealed that this procedure is sufficient to replace histology in polyp surveillance. This was a large observational study investigating the feasibility of the “resect and discard” strategy among nonaffiliated hospitals across China. Subjects achieved a 93.2% NPV for rectosigmoid polyps with HC profiling, successfully exceeding the 90% threshold proposed by the ASGE.[1] Upon integration of histology-based LC profiling and polyps >5 mm with NBI-based estimation of HC diminutive polyps, the overall surveillance interval agreement was 91.7%. In terms of incorrect surveillance interval estimates, 21.1% led to earlier than recommended surveillance colonoscopies, ensuring that 93.5% of patients received timely (91.7%) or early (1.8%) surveillance procedures. However, it should be noted that the absolute number of small polyps is very high, representing an overwhelming proportion of all polyps. Therefore, although the proportion of missed diagnoses of pathological features is low, when multiplied by the large absolute number of small polyps, their contributions to the overall number of missed diagnoses cannot be ignored, eventually contributing to the burden of colorectal cancer. In conclusion, with routine training, gastroenterologists without earlier NBI training from nonaffiliated hospitals across China satisfied the ≥90% NPV and ≥90% surveillance interval thresholds recommended by the ASGE. Funding This work was supported by grants from the Science and Technology Commission of Shanghai Municipality (No. 23Y11905300), Noncommunicable Chronic Diseases–National Science and Technology Major Project (No. 2024ZD0520800), and Fundamental Research Funds for the Central Universities (No. YG2025ZD01). Conflicts of interest None.