If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Gastroenterology and Interventional Endoscopy, Krankenhaus Barmherzige Brüder (St. John of God Hospital), Teaching Hospital of the University of Regensburg, Regensburg, Germany
The Narrow-band Imaging International Colorectal Endoscopic (NICE) Classification has been validated for differentiating hyperplastic from adenomatous polyps. This classification system was based on narrow-band imaging (NBI) technology, leaving uncertainty regarding its applicability to other systems. The aim of this study was to assess accuracy and reliability of histologic predictions for polyps <1 cm by applying the NICE classification to the Fujinon Spectral Imaging Color Enhancement (FICE) System.
Methods
A video library of 55 polyps <1 cm histologically verified with FICE was prospectively created, including polyps that fulfilled inclusion criteria (morphology, size, histology) in consecutive colonoscopies. Six endoscopists with experience in electronic chromoendoscopy independently reviewed the polyp images, scored the polyps as adenomatous or hyperplastic, and assigned a level of confidence to the predictions. Twenty videos were reassessed at 6 months. The diagnostic performances of the endoscopists was calculated both combined and individually according to the histopathology of the polyps. A mixed-effect logistic regression model, in which polyps were considered as random effects, and polyp histology, confidence level, and readers were considered as fixed effects, was used. Results were expressed as odds ratios (ORs) with 95% confidence intervals (CIs).
Results
Of the 55 polyps (mean size 4.6 mm), 29 (53%) were adenomas, and 26 (47%) were hyperplastic. Across all the readers and observations, the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and area under the curve (AUC) were 77%, 75%, 88%, 75%, 77%, and 0.82, respectively. Individual rater accuracy ranged from 66% to 96%, being <90% in 5 of 6 cases. Overall, 68.5% of predictions (226/330) were made with high confidence, although there was high variability (Fleiss kappa, 0.15; 95% CI, 0.08-0.22). Sensitivity, specificity, PPV, NPV, accuracy, and AUC for predictions made with high confidence were 81%, 80.5%, 80%, 77%, 82%, and 0.88 being significantly more accurate as compared with a low confidence of diagnosis (OR 2.4; 95% CI, 1.2-4.7). Regarding the performance of the individual NICE criteria, the odds of adenoma detection were 3.4 (95% CI, 1.8-6.3) and 4.0 (95% CI, 2.1-7.5) by using surface and vessels patterns alone, as compared with the color criterion. Interrater and intrarater agreement with the NICE was only moderate (interrater: Fleiss kappa, 0.51; 95% CI, 0.44-0.56; intrarater: kappa, 0.40; 95% CI, 0.20-0.60).
Conclusions
The application of the NICE classification to FICE resulted in suboptimal accuracy and only moderate interobserver agreement.
Colorectal cancer (CRC) represents an important health problem in Western countries. In Europe, almost 413,000 people are newly diagnosed with CRC, and about half of them will die of the disease, CRC representing the second leading cause of cancer deaths.
Screening and surveillance for the early detection of colorectal cancer and adenomatous polyps, 2008: a joint guideline from the American Cancer Society, the US Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology.
On the other hand, the removal of hyperplastic polyps generally has been regarded as a false-positive result of this procedure, except for larger hyperplastic or other serrated lesions. Because of the impossibility of predicting final histology with white-light endoscopy, endoscopists remove all the detected polyps in order to refer them to pathology. However, the cost of pathology examination for small hyperplastic polyps is likely to negatively affect the cost effectiveness of the procedure.
The possibility of harboring a polyp with hyperplastic histology is much higher in the small (<10 mm) than in the large lesions, and it is also higher in diminutive (≤5 mm) than in 6- to 9-mm polyps. Of note, diminutive and small polyps account for over 80% of all the polypoid lesions.
Chromoendoscopy using indigo carmine dye spraying with magnifying observation is the most reliable method for differential diagnosis between non-neoplastic and neoplastic colorectal lesions: a prospective study.
was revolutionized by the development of electronic or virtual chromoendoscopy. The main advantages of electronic or virtual chromoendoscopy are the simple and immediate activation and the wide availability on the new generations of endoscopes. In order to differentiate between neoplastic (adenomatous) and nonneoplastic (hyperplastic) lesions, electronic or virtual chromoendoscopy exploits the neoangiogenesis of neoplastic lesions as well as the mucosal pit pattern. Promising results in differentiating between different histologies recently have been reported in a meta-analysis with all of the available technologies.
Diagnostic performance of narrowed spectrum endoscopy, autofluorescence imaging, and confocal laser endomicroscopy for optical diagnosis of colonic polyps: a meta-analysis.
This has prompted the development of the resect-and-discard strategies, with no resection and/or histopathology assessment for clinically irrelevant diminutive lesions.
The American Society for Gastrointestinal Endoscopy PIVI (Preservation and Incorporation of Valuable Endoscopic Innovations) on real-time endoscopic assessment of the histology of diminutive colorectal polyps.
Clinical application of narrow-band imaging (NBI) in particular has been standardized by the validation of the Narrow-band Imaging International Colorectal Endoscopic (NICE) Classification.
In contrast, Fujinon Spectral Imaging Color Enhancement (FICE, Fujifilm Europe GmbH, Duesseldorf, Germany) is a postprocessor technology that electronically removes the red part of the waveband in order to achieve vascular enhancement.
Because of the underlying differences between the NBI and FICE technologies, it remains uncertain whether the NICE classification may be translated to FICE.
The aim of our study was to validate the NICE classification for differentiating hyperplastic and adenomatous polyps <1 cm when diagnosed with high confidence by using FICE with high definition without optical magnification: the REDEFINE (REsect and Discard study Extension to FIce of NIce) study.
Methods
A video library of polyps <10 mm characterized with FICE technology was prospectively created for the purpose of this study. Six experienced endoscopists applied the NICE classification to these videos, and their performance in terms of diagnostic confidence and accuracy was measured. All institutions participating in this noninterventional clinical study obtained the appropriate institutional review board approval.
Study population
Consecutive adult patients referred for elective outpatient colonoscopy were enrolled from June to November 2014. Exclusion criteria were inflammatory bowel disease, a personal history of polyposis syndrome, diverticulitis or toxic megacolon, and a history of radiation therapy to the abdomen or pelvis. Patients with histories of severe cardiovascular, pulmonary, liver, or renal disease as well as those with coagulation disorders or use of anticoagulants were excluded.
Video library
One expert endoscopist who was not involved in the subsequent phases of the study (C.C.) recorded high-definition videos of consecutive polyps <1 cm diagnosed in the study period. All endoscopies were performed with a Fujinon colonoscope series EC-600 with the FICE system (VD-4450-HD, FUJIFILM Europe GmbH, Duesseldorf, Germany). Each video consisted of 5 to 10 seconds of white-light endoscopy and 30 to 60 seconds of FICE use (types 1, 4, and 8). For the purpose of this study, magnification was not allowed during recording. All polyps were resected and sent for histopathologic examination that was used as the criterion standard for our analysis, and that was performed by an experienced GI pathologist according to the revised Vienna classification.
we decided not to include them in the present analysis.
Endoscopist reading
Endoscopists with at least 2 years of regular practice with advanced endoscopic imaging, such as NBI or FICE, were selected. After a formal session on the NICE classification based on its theoretical background and a series of NBI-based cases,
Accuracy of narrow-band imaging in predicting colonoscopy surveillance intervals and histology of distal diminutive polyps: results from a multicenter, prospective trial.
In detail, each endoscopist individually scored each of the 3 criteria of the NICE system (color/vessel/surface pattern) as present/absent/uncertain and eventually indicated the overall level of confidence (high/low/no confidence) and the possible NICE category, that is, type 1 for nonadenomatous and type 2 for adenomatous, in a dedicated form (Supplemental Table 1, available online at www.giejournal.org).
Table 1The Narrow-band Imaging International Colorectal Endoscopic Classification
Polyps were predicted to be adenomatous or hyperplastic with high confidence if they had one or more features associated with one histology and no features associated with the other histology. If there was uncertainty regarding the features or if there were features of both adenomatous and hyperplastic polyps, then a prediction of histology was made but with low confidence.
Reading of 20 randomly selected videos was thereafter repeated after 6 months by the same observers to assess the intraobserver agreement.
Study outcomes
Sensitivity and specificity of FICE characterization were calculated as the rate of histologically verified <10 mm adenomas and hyperplastic polyps, respectively, that were correctly classified by the in vivo FICE diagnosis. Intraobserver concordance was assessed by comparing the accuracy values of the first 55 with those of the 20 randomly selected for the 6-month reading.
Primary outcomes
Primary outcomes included diagnostic accuracy of FICE for differentiating between the nonadenomatous and adenomatous histotypes in diagnoses with high confidence.
Secondary outcomes
Secondary outcomes were to assess (1) interobserver and intraobserver agreement, (2) diagnostic accuracy of the individual criteria of the NICE classification, and (3) diagnostic accuracy according to the level of confidence.
The American Society for Gastrointestinal Endoscopy PIVI (Preservation and Incorporation of Valuable Endoscopic Innovations) on real-time endoscopic assessment of the histology of diminutive colorectal polyps.
it was assumed that the overall accuracy for classifying histology of polyps scored with high confidence was 90% and that 80% of predictions would be made with high confidence. With 6 observers and at least 40 polyps (240 polyps observation = 40 polyps × 6 observers), the study could estimate a 95% confidence interval (CI) of the accuracy with a ±5% margin of error. The sample size was increased to 330 observations (55 polyps × 6 observers) to allow for nonindependence of data by using more than 1 observer. Furthermore, assuming that 2 observers would agree about 50% of the observations, a total of 55 polyps ensured estimation of the agreement (measured by k statistic) of each pair of observers with a margin of error of ±7%.
Statistical analysis
Diagnostic values of each criterion used alone were assessed. The presence of an adenomatous feature at each criterion was defined as a positive result. Sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV), and area under the receiver operating characteristic curves (AUC) were calculated along with 95% CIs. A bivariate model to account for within-observer correlation was used to estimate diagnostic characteristics pooled across the study readers.
Robust standard errors also were used for all analyses to account for the fact that each polyp was included in the analysis 6 times (ie, 1 for each observer).
The sensitivity and specificity levels of the different criteria were compared by using multilevel logistic regression. Polyps and observers were considered as random cross-classification, with polyp characteristics (ie, polyp size, polyp location in the colon, morphology, and histology) and the criterion as random fixed effects. The effect of the criterion and polyp characteristics on the odds of adenoma (hyperplastic polyps) detection was assessed by using a Wald test or joint Wald test as appropriate. Tests of interaction assessed whether there was any important effect modification, for example, whether the effect of 1 factor (such as classification criterion) changed for differing levels of other factors (such as polyp histology). Data were presented as odds ratios (ORs) and 95% CIs. The OR is interpreted as the ratio of the odds of adenoma (hyperplastic polyps) detection for 1 group (eg, color criterion used alone) or circumstance compared with the odds for another (eg, vessels criterion used alone). Finally, diagnostic values of the criteria used in combination (combination of at least 1 criterion being positive vs all combined criteria being positive) were assessed and compared by using the methodology described earlier. All analyses were conducted in R Version 3.8 (R Core Team [2015]. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/) program.
Results
Creation of FICE cases video library
Overall, 55 polyps <10 mm removed in 30 patients were included. All polyps were recorded at both white-light and FICE, and were sent for histology. In detail, 29 polyps (53%) were adenomatous, and 26 (47%) were hyperplastic. Regarding the dimension, median size was 4.6 mm (3-8 mm), and 85% of the lesions were diminutive (≤5 mm). In addition, 65% of the lesions were sessile, and the remaining were nonpolypoid (II-a according to Paris classification). Regarding the location, 22 polyps (44%) were located in the rectosigmoid tract, whereas the others were more proximal (Table 2).
Table 2Distribution of the <10 mm polyps included in the video library
After the first review of the video library, 330 (55 polyps × 6) observations were analyzed. When not considering the level of confidence, the overall sensitivity, specificity, positive and negative predictive values, accuracy, and AUC for diagnosing adenomas were 77% (95% CI, 69%-83%), 75% (95% CI, 67%-82%), 88% (95% CI, 80%-92%), 75% (95% CI, 67%-82%), 77% (95% CI, 67%-83%), and 0.82 (95% CI, 0.67-0.89), respectively, with a moderate interobserver agreement (Fleiss kappa, 0.51; 95% CI, 0.44-0.56) (Table 3). Individual observer accuracy ranged from 66% to 96%.
Table 3Accuracy values according to individual readers
When classifying the 330 observations according to the level of confidence (Table 4), we made the diagnosis with high confidence in 68.5% (226/330), although there was high variability among the observers (Fleiss kappa 0.15; 95% CI, 0.08-0.22). Sensitivity, specificity, PPV, NPV, accuracy, and AUC for predictions made with high and low confidence were 81% (95% CI, 65%-90%), 80.5% (95% CI, 62%-91%), 80% (95% CI, 71%-87%), 77% (95% CI, 67%-84%), 82% (95% CI, 76%-86%), 0.88 (95% CI, 0.83-0.94), and 58% (95% CI, 42%-70%), 69% (95% CI, 51%-82%), 66% (95% CI, 51%-79%), 56% (95% CI, 41%-70%), 66% (95% CI, 56%-75%), and 0.76 (95% CI, 0.67-0.88), respectively.
Table 4Diagnostic accuracy according to the level of confidence
Regular experience with Fujinon Spectral Imaging Color Enhancement only.
0.77 (0.61-0.88)
0.57 (0.29-0.82)
∗ Regular (>2 y) experience with Narrow-band Imaging International Colorectal Endoscopic Classification and Fujinon Spectral Imaging Color Enhancement.
† Regular experience with Narrow-band Imaging International Colorectal Endoscopic Classification only.
‡ Regular experience with Fujinon Spectral Imaging Color Enhancement only.
When we adjusted the level of diagnosis for potential confounders at multivariate analysis, a high confidence of diagnosis was significantly more likely to be correct compared with a low confidence diagnosis (OR 2.4; 95% CI, 1.2-4.7; P = .015). Correct histologic prediction by FICE did not appear to be related to the histologic type of the polyps (adenomatous vs hyperplastic polyp: OR 0.7; 95% CI, 0.21-2.4; P = .568).
Reading accuracy according to the NICE individual criteria
Table 5 shows the diagnostic values of the criteria, used alone and in combination, for each observer. Across all polyps and study observers, sensitivity and specificity of the color criterion alone ranged from 40% to 90% (pooled estimate 60%; 95% CI, 45%-73%; P = .020) and from 64% to 92% (pooled estimate 79%; 95% CI, 70%-86%; P = .321), corresponding PPV and NPV being 77% (95% CI, 69%-83%) and 62% (95% CI, 53%-70%). Accuracy of vessels criterion alone varied, with sensitivity levels of 55% to 97% (pooled estimate 74%; 95% CI, 61%-84%; P = .030) and specificity levels of 64% to 92% (pooled estimate 78%, 95% CI, 70%-84%; P = .151). The pooled PPV and NPV of this criterion were 80% (95% CI, 73%-86%) and 71% (95% CI, 63%-77%), respectively. For surface criterion alone, sensitivity was between 55% and 90% (P = .031; pooled sensitivity 74%; 95% CI, 62%-83%) and specificity between 64% and 92% (P = .197; pooled specificity 80%; 95% CI, 70%-87%), resulting in a PPV of 81% (95% CI, 73%-86%) and an NPV of 71% (95% CI, 63%-78%). Color pattern alone had lower sensitivity than vessels and surface pattern (60% vs 74%; P < .001), whereas specificity was similar (79% vs 78% vs vessels pattern and 79% vs 80% vssurface pattern). In combination, when at least 1 of the 2 criteria being positive was used, the specificity of either the surface or vessels pattern being positive was higher than that achieved with either color or vessels pattern (74% vs 67%; P = .041), whereas sensitivity was similar (P = 1.000). When both criteria being positive were used, the sensitivity of both surface and vessels patterns, compared with both color and surface patterns, was significantly higher (74% vs 55%; P < .001), although specificity was lower (81% vs 90%).
Table 5Performance characteristics of the NBI criteria used in isolation and in combination
All 2-way interactions between polyp size, location in the colon, polyp morphology, and criterion were not significant and therefore were not included in the final model. According to the results from the regression model, the effect of the criterion on adenoma detection was significant (P < .001). Compared with using the color pattern alone, the odds of adenoma detection were 3.4 (95% CI, 1.8-6.3) and 4.0 (95% CI, 2.1-7.5) by using surface and vessels pattern alone, respectively (Videos 1 and 2, available online at www.giejournal.org). The effect of other factors such as polyp morphology (flat versus sessile, OR 0.3; 95% CI, 0.03-3.0; P = .303), polyp size (OR 0.9; 95% CI, 0.4-2.1; P = .725), and polyp location (proximal vs left side of the colon: OR 0.4; 95% CI, 0.03-4.9; P = .448) on the diagnostic accuracy for adenoma was not significant. Differently, the criterion used did not significantly influence specificity. Compared with using color pattern alone, the odds of hyperplastic polyp detection were 1.1 (95% CI, 0.6-2.0) and 0.9 (95% CI, 0.5-1.8) when we used surface and vessels pattern alone, respectively. Polyp morphology (flat versus sessile, OR 1.0; 95% CI, 0.3-3.9; P = .991) and size (OR 0.8, 95% CI, 0.4-1.2; P = .362) did not influence hyperplastic polyp detection. The only factor found to influence specificity was colon polyp location. Independently from the criterion used, hyperplastic polyps located distally were less likely to be correctly detected than those located proximally (OR 0.1; 95% CI, 0.02-0.7; P = .017).
Table 5 also presents the diagnostic values for combinations of different criteria. At multivariate analysis, the combination of 1 of the 2 criteria being positive compared with that of both criteria being positive significantly increased adenoma detection (OR 4.9; 95% CI, 3.3-7.4; P < .001), although it decreased specificity (OR 0.18; 95% CI, 0.11-0.28; 95%; P < .001). When we compared different combinations of at least 1 criterion being positive, there was some evidence that the presence of an adenomatous feature at either surface or vessels pattern increased specificity; the odds of hyperplastic polyp detection by using either the surface or vessels pattern was 1.8 (95% CI, 1.0-3.4; P = .051) times greater than those by using either color or vessels pattern. Furthermore, the combination with either surface or vessels pattern being positive, compared with that with either color or vessels pattern, had similar sensitivity (OR 0.85; 95% CI, 0.3-1.8; P = .514). When comparing combinations of both criteria being positive, the presence of adenomatous features at both surface and vessels criteria significantly increased adenoma detection: the odds of adenoma detection by using both surface and vessels were 3.3 (95% CI, 1.8-6.0; P < .001) compared with using both color and vessels. However, the combination of both surface and vessels pattern being positive significantly decreased specificity: OR 0.11; 95% CI, 0.04-0.32; P < .001. Corresponding values for each reader are reported in Supplemental Table 2 (available online at www.giejournal.org).
Intraobserver agreement
Of the 20 polyps randomly selected for the second reading at 6 months from the index observation, 10 (50%) were adenomas, and 10 (50%) were hyperplastic polyps (Table 6). Overall, observers assigned 76% of the observations to the same rating. However, the overall agreement between the first and the second evaluations was moderate (k = 0.40; 95% CI, 0.20-0.60) (Fig. 1). Among observers, kappa ranged from 0.22 to 0.50. The overall intraobserver agreement for FICE feasibility of adenomas was 0.60 (95% CI, 0.35-0.84), whereas, for hyperplastic polyps it was 0.21 (95% CI, 0-0.50). The intraobserver agreement for classification of the specific histologic type of the 20 lesions by adenoma or hyperplastic polyp ranged from 55% to 95%; this corresponded to a kappa range of 0.4 to 0.90. Overall, the percentage of agreement between the first and the second evaluation was 0.69 (95% CI, 0.56-0.80; kappa = 0.32; 95% CI, 0.08-0.58) for adenomas and 0.80 for hyperplastic polyps (95% CI, 0.66-0.88; kappa = 0.25; 95% CI, 0-0.62).
Table 6Cross tabulations of FCI feasibility scores by the same reader at 2 different time points
Among all 120 polyp observations (6 observers × 20 polyps = 120 observations), there were 5 missing data. Overall, intraobserver agreement between the first and second evaluations was moderate: kappa = 0.40 (95% CI, 0.20-0.60). Among observers, kappa ranged from 0.22 (fair agreement) to 0.50 (moderate agreement).
FICE, Fujinon spectral imaging color enhancement.
∗ Among all 120 polyp observations (6 observers × 20 polyps = 120 observations), there were 5 missing data. Overall, intraobserver agreement between the first and second evaluations was moderate: kappa = 0.40 (95% CI, 0.20-0.60). Among observers, kappa ranged from 0.22 (fair agreement) to 0.50 (moderate agreement).
The results of our study do not support the application of the NICE classification to the FICE technology, because of the overall low accuracy coupled with a high interobserver and intraobserver variability. However, our data confirm the association between the level of diagnostic confidence and the overall accuracy already shown with the NICE classification.
These data are relevant for the following reasons. First, we showed that a classification based on 1 technology cannot be immediately applied to different technologies. In detail, our overall accuracy of 77% was much lower than the >90% shown in the validation of the NICE classification with NBI technology, with 5 of the 6 included endoscopists failing to reach the 90% threshold.
When we considered the >2-year experience in advanced endoscopic imaging of the participating endoscopists, the suboptimal results of our study may be attributed at least in part to the different filtering of the light between the 2 systems, that is, preprocessing physical filter with NBI and postprocessing elaboration with FICE. This is corroborated with the similar performance of the endoscopists with regular experience in NBI or FICE previous to the study. Second, the moderate interobserver agreement (k = 0.51) observed in our study did not confirm the corresponding high value (k = 0.87) of the original NICE publication,
indicating that the same NICE criteria were interpreted differently across the study endoscopists, questioning the reproducibility of such classification when applied to the FICE system. It is unlikely that a short training course with pathology-controlled cases would have achieved a high accuracy with the FICE system because it was reported for NBI.
Third, the intraobserver agreement was only moderate, being particularly low for adenomas. This would indicate an endoscopist-related variability when the same classification was applied to similar cases.
In our study, a high confidence of diagnosis was reached only in 68.5% of cases, which appears substantially lower than the 80% reported in the NICE validation study.
This is likely to be related to the subjective perception by the endoscopists of the reduced accuracy of the NICE classification with a different technology. Although this does not directly affect the clinical performance of the classification, such a low rate of high-confidence diagnoses would reduce the absolute savings related with resect-and-discard strategies because all polyps characterized with a low level of confidence should be sent for histologic examination. Of note, the results of our multivariate analysis confirm the strict association between the level of diagnostic confidence and the accuracy of the NICE classification.
Ladabaum U, Fioritto A, Mitani A, et al. Real-time optical biopsy of colon polyps with narrow band imaging in community practice does not yet meet key thresholds for clinical decisions. Gastroenterology. Epub 2012 Oct 3.
In particular, the overall accuracy improved from 77% to 82%, with 3 readers presenting a ≥85% accuracy. Despite the fact that these values are much lower than the >98% values shown when the NBI technology was used,
our data would support an independence between the association level of diagnosis accuracy and the technology adopted.
When the performance of the individual criteria of the NICE classification was explored, the color criteria appeared to be the main factor responsible for the suboptimal performance shown in our study. It was associated with a substantially lower sensitivity and not superior specificity, as compared with the vessel and surface criteria. This would indicate that a substantially lower number of adenomas failed to appear browner relative to the background with FICE, despite presenting with a surface and/or vessel pattern compatible with an adenomatous nature. Moreover, the lack of an increased specificity with the color criterion would suggest that such a criterion was somewhat down-staged as compared with the other 2 criteria when predicting the hyperplastic nature of the examined polyps. The suboptimal performance of the color criterion with the FICE system would suggest a less chromatic contrast between the detected lesion and the surrounding mucosa, as compared with the NBI technology.
The main strengths of our analysis are related to the pure application of the NICE classification to the FICE system. In fact, if we had developed a new FICE-based classification, there would have been residual uncertainty reagarding whether the NICE classification was applicable to different technologies. Second, unlike the initial validation of the NICE classification, we preferred short videos rather than still images because they are more comparable to actual clinical practice. Indeed, in the NICE validation, the actual performance in a real clinical scenario was substantially worse than that reported for still images (89% vs 99%), although still superior to the suboptimal performance reported in our study. Readers were not required to rate the quality of the videos. Thus, we cannot exclude that some videos would have been rated as suboptimal by 1 or more raters. However, it is likely that such video rating would have affected the ratio between low-confidence and high-confidence predictions more, rather than the intrinsic accuracy of the high-confidence diagnosis that was the primary end-point of our diagnosis. In addition, our data are in line with suboptimal performance reported in a community-based study.
Ladabaum U, Fioritto A, Mitani A, et al. Real-time optical biopsy of colon polyps with narrow band imaging in community practice does not yet meet key thresholds for clinical decisions. Gastroenterology. Epub 2012 Oct 3.
Third, because there is uncertainty regarding the most suitable setting for polyp characterization, we applied 3 different FICE settings to the same lesion, in order to take into account different enhancements of the vessel and surface patterns.
The main limitation of our analysis is that we did not perform a head-to-head comparison between the NBI and the FICE technologies. However, from a clinical perspective, this is not practical. Second, we did not apply the NICE classification with the FICE system in real-time colonoscopy, so that we did not formally assess the benchmark provided by the American Society for Gastrointestinal Endoscopy for the resect-and-discard strategies, including the rate of patients with an adequately assigned postpolypectomy surveillance interval and the NPV for an adenomatous histology for diminutive rectosigmoid polyps.
The American Society for Gastrointestinal Endoscopy PIVI (Preservation and Incorporation of Valuable Endoscopic Innovations) on real-time endoscopic assessment of the histology of diminutive colorectal polyps.
Accuracy of narrow-band imaging in predicting colonoscopy surveillance intervals and histology of distal diminutive polyps: results from a multicenter, prospective trial.
However, it is implausible that any of the 2 main benchmarks would have been reached, when the very low accuracy shown in our analysis is considered. Third, we decided to exclude sessile serrated polyps from our analysis, because it was initially done in the NICE validation process. In fact, a new classification, not purely based on advanced endoscopic imaging, has been validated to characterize sessile serrated polyps.
Dutch Workgroup serrAted polypS & Polyposis (WASP). Development and validation of the WASP classification system for optical diagnosis of adenomas, hyperplastic polyps and sessile serrated adenomas/polyps.
Gut.2015 Mar 9; (pii: gutjnl-2014-308411. doi: 10.1136/gutjnl-2014-308411. [Epub ahead of print])
The American Society for Gastrointestinal Endoscopy PIVI (Preservation and Incorporation of Valuable Endoscopic Innovations) on real-time endoscopic assessment of the histology of diminutive colorectal polyps.
so we cannot exclude that its addition would result in incremental accuracy. For instance, a similar performance between magnifying NBI and FICE for prediction of submucosal invasion was shown in a retrospective single-center Japanese study.
Fifth, as already mentioned, we did not intend to develop a new FICE-based classification. The development of a new classification should be based on additional data as well as on other factors, such as the availability of the FICE system and the potential savings related to the resect-and-discard policy. Sixth, our sample size was designed on the expected accuracy for histology prediction of the polyp, similarly to the initial report of the NICE validation.
The American Society for Gastrointestinal Endoscopy PIVI (Preservation and Incorporation of Valuable Endoscopic Innovations) on real-time endoscopic assessment of the histology of diminutive colorectal polyps.
Thus, although we found significant associations between individual NICE criteria and FICE performances, larger studies are needed to further explore the actual role of each of the NICE criteria alone or in combination with FICE predictions. Seventh, our estimate of the intraobserver agreement could have been affected by the technology used in the 6-month period between the 2 ratings. In particular, those endoscopists who used FICE could have maintained a higher level of competency because of the additional exposure to this technology. Eighth, we did not select our reading endoscopists according to a posttraining examination in order not to artificially enrich our study population of high-performing operators. Thus, we cannot exclude that a posttraining selection would not have resulted in better results.
In conclusion, our analysis showed a suboptimal performance when applying the NICE classification to the FICE system, suggesting caution when applying the NICE criteria to other technological platforms without previous validation.
Appendix
Supplemental Table 1Form to be filled out by each endoscopist after reviewing each case
Physician name:
Endoscopic experience, yr:
FICE/NBI experience, yr:
Case no:
Was white-light characterization feasible?
□ Yes, with high confidence□ Yes, with low confidence □ No
Was FICE characterization feasible?
□ Yes, with high confidence□ Yes, with low confidence □ No
Which were the FICE characteristics of the polyp?
Color: (level of confidence: □ high □ low) □ Same or lighter than background □ Darker relative to background (verify color arises from vessels)
Vessels: (level of confidence: □ high □ low) □ None or isolated lacy vessels coursing across the lesion □ Brown vessels surrounding white structures
Surface pattern: (level of confidence: □ high □ low) □ Brown or white spots of uniform size □ Homogeneous absence of pattern □ Oval white structures surrounded by brown vessels □ Tubular white structures surrounded by brown vessels □ Blanched white structures surrounded by brown vessels □ Elongated white structures surrounded by brown vessels
Was the pit pattern characterization feasible? □ Yes, with high confidence □ Yes, with low confidence □ No If yes, pit pattern: □ I-II □ III or other
What was the type of polyp according to FICE diagnosis? □ Hyperplastic (type 1) □ Adenomatous (type 2) □ Both hyperplastic and adenomatous □ Other ____________________________
FICE, Fujinon Spectral Imaging Color Enhancement; NBI, Narrow-band Imaging International Colorectal Endoscopic Classification.
Supplemental Table 2Performance characteristics of the Narrow-band Imaging International Colorectal Endoscopic Criteria (used alone) across the study readers for each individual criterion (A) or the possible combinations (B)
(A) Performance indicator (%)
1
2
3
4
5
6
All observers
Color criterion (stand-alone diagnostic performance)
Screening and surveillance for the early detection of colorectal cancer and adenomatous polyps, 2008: a joint guideline from the American Cancer Society, the US Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology.
Chromoendoscopy using indigo carmine dye spraying with magnifying observation is the most reliable method for differential diagnosis between non-neoplastic and neoplastic colorectal lesions: a prospective study.
Diagnostic performance of narrowed spectrum endoscopy, autofluorescence imaging, and confocal laser endomicroscopy for optical diagnosis of colonic polyps: a meta-analysis.
The American Society for Gastrointestinal Endoscopy PIVI (Preservation and Incorporation of Valuable Endoscopic Innovations) on real-time endoscopic assessment of the histology of diminutive colorectal polyps.
Accuracy of narrow-band imaging in predicting colonoscopy surveillance intervals and histology of distal diminutive polyps: results from a multicenter, prospective trial.
Ladabaum U, Fioritto A, Mitani A, et al. Real-time optical biopsy of colon polyps with narrow band imaging in community practice does not yet meet key thresholds for clinical decisions. Gastroenterology. Epub 2012 Oct 3.
Dutch Workgroup serrAted polypS & Polyposis (WASP). Development and validation of the WASP classification system for optical diagnosis of adenomas, hyperplastic polyps and sessile serrated adenomas/polyps.
Gut.2015 Mar 9; (pii: gutjnl-2014-308411. doi: 10.1136/gutjnl-2014-308411. [Epub ahead of print])
DISCLOSURE: R. Bisschops received a speaker’s fee from Pentax, Olympus, Fujifilm, research support from Olympus, Pentax, and is supported by a grant from the Research Foundation Flanders (FWO-Vlaanderen). E. Dekker received research support, equipment on loan, and a speaker’s fee from Olympus and equipment on loan from Fujinon as well as a speaker’s fee from Tillots. C. Hassan received equipment on loan from Fujinon, and Fujinon provided equipment. All other authors disclosed no financial relationships relevant to this publication.
If you would like to chat with an author of this article, you may contact Dr Repici at [email protected]
00170-X&id=gr1.jpg){kind=link}