- Email: [email protected]
Endoscopic prediction of deep submucosal invasive carcinoma: validation of the Narrow-Band Imaging International Colorectal Endoscopic (NICE) classification
ORIGINAL ARTICLE: Clinical Endoscopy
Endoscopic prediction of deep submucosal invasive carcinoma: validation of the Narrow-Band Imaging International Colorectal Endoscopic (NICE) classification Nana Hayashi, MD,1 Shinji Tanaka, MD, PhD,1 David G. Hewett, MD, PhD,2 Tonya R. Kaltenbach, MD, PhD,3 Yasushi Sano, MD, PhD,4 Thierry Ponchon, MD, PhD,5 Brian P. Saunders, MD, PhD,6 Douglas K. Rex, MD, PhD,7 Roy M. Soetikno, MD, PhD3 Hiroshima, Japan
Background: A simple endoscopic classification to accurately predict deep submucosal invasive (SM-d) carcinoma would be clinically useful. Objective: To develop and assess the validity of the NBI international colorectal endoscopic (NICE) classification for the characterization of SM-d carcinoma. Design: The study was conducted in 4 phases: (1) evaluation of endoscopic differentiation by NBI-experienced colonoscopists; (2) extension of the NICE classification to incorporate SM-d (type 3) by using a modified Delphi method; (3) prospective validation of the individual criteria by inexperienced participants, by using highdefinition still images without magnification of known histology; and (4) prospective validation of the individual criteria and overall classification by inexperienced participants after training. Setting: Japanese academic unit. Main Outcome Measurements: Performance characteristics of the NICE criteria (phase 3) and overall classification (phase 4) for SM-d carcinoma; sensitivity, specificity, predictive values, and accuracy. Results: We expanded the NICE classification for the endoscopic diagnosis of SM-d carcinoma (type 3) and established the predictive validity of its individual components. The negative predictive values of the individual criteria for diagnosis of SM-d carcinoma were 76.2% (color), 88.5% (vessels), and 79.1% (surface pattern). When any 1 of the 3 SM-d criteria was present, the sensitivity was 94.9%, and the negative predictive value was 95.9%. The overall sensitivity and negative predictive value of a global, high-confidence prediction of SM-d carcinoma was 92%. Interobserver agreement for an overall SM-d carcinoma prediction was substantial (kappa 0.70). Limitations: Single Japanese center, use of still images without prospective clinical evaluation. Conclusion: The NICE classification is a valid tool for predicting SM-d carcinomas in colorectal tumors. (Gastrointest Endosc 2013;78:625-32.)
Abbreviations: NBI, narrow-band imaging; NICE, NBI international colorectal endoscopic; SM-d, deep submucosal invasion. DISCLOSURE: All authors disclosed no financial relationships relevant to this publication. See CME section; p. 637.
Use your mobile device to scan this QR code and watch the author interview. Download a free QR code scanner by searching “QR Scanner” in your mobile device’s app store. www.giejournal.org
Copyright ª 2013 by the American Society for Gastrointestinal Endoscopy 0016-5107/$36.00 http://dx.doi.org/10.1016/j.gie.2013.04.185 Received October 9, 2012. Accepted April 20, 2013. Current affiliations: Department of Endoscopy, Hiroshima University Hospital, Hiroshima, Japan (1), University of Queensland School of Medicine, Australia (2), Veterans Affairs Palo Alto Health Care System, Palo Alto, California (3), Sano Hospital, Hyogo, Japan (4), Hôpital Edouard-Herriot, Lyon, France (5), St. Mark’s Hospital, London, United Kingdom (6), Indiana University, Indianapolis, Indiana, USA (7). Reprint requests: Shinji Tanaka, MD, PhD, Department of Endoscopy, Hiroshima University, 1-2-3 Kasumi, Minami-ku, Hiroshima 734-8551, Japan.
Volume 78, No. 4 : 2013 GASTROINTESTINAL ENDOSCOPY 625
Validation of the NICE classification
The potential for endoscopic treatment of early colorectal cancer is being realized through recent advances in endoscopic imaging and resection techniques. These resections require the ability to select the appropriate neoplastic lesions because patients with high-grade dysplasia and submucosal invasive carcinoma with an invasion depth of !1000 mm (superficial submucosal invasion), absence of tumor budding, and absence of lymphovascular invasion are likely to be cured after endoscopic resection,1-3 whereas those with deeper submucosal invasion have up to a 10% risk of lymph node metastasis.1,2 The endoscopic resection of deep (R1000 mm) submucosal invasive (SM-d) carcinoma, also is associated with an increased risk of bleeding and perforation.4 Clinical experience with detailed inspection of the continuous spectrum of adenomas as they progress to carcinoma suggests that their appearance under narrow-band imaging (NBI) can be artificially classified into 2 typesd type 2 and type 3.5 Type 3 includes SM-d carcinoma, whereas type 2 includes adenomas to superficial invasive carcinoma. A number of studies have suggested the potential validity of these methods of classification,6-36 but the validity of their component criteria has not been established. In this study, we evaluated the predictive validity of an expanded classification for the endoscopic differentiation of deep submucosal colorectal carcinoma from tubular adenomas, adenomas with high-grade dysplasia, and superficial submucosal carcinoma. Specifically, we aimed to (1) expand the NBI international colorectal endoscopic (NICE) classification (Fig. 1) to include SM-d carcinoma (type 3), in addition to type 1 (hyperplastic) and type 2 (adenomatous to superficial submucosal colorectal carcinoma)5,7,37; (2) evaluate the predictive validity of the components of the classification; and (3) evaluate the validity and performance characteristics of the overall classification for distinguishing type 3 from types 1 and 2.
METHODS Study design We conducted a prospective validation study of the NICE classification for the diagnosis of SM-d carcinoma in colorectal tumors by using NBI without optical magnification. Our approach here reflects the one used in our previous validation study of classifying type 1 versus type 2.37 Specifically, the study was conducted in 4 phases (Fig. 2): (1) evaluation of the accuracy of NBI-experienced colonoscopists in predicting SM-d, (2) development of an expanded NICE classification to incorporate a further category (type 3) for SM-d, (3) validation of the component criteria by testing the criteria in endoscopy-naïve participants, and (4) assessment of the performance and reliability of the overall classification for predicting histology in participants trained in NBI diagnosis. The Institutional Review Board at Hiroshima University Hospital approved the study. 626 GASTROINTESTINAL ENDOSCOPY Volume 78, No. 4 : 2013
Hayashi et al
Take-home message The authors aimed to establish the predictive validity of simple classification for the endoscopic differentiation of deep submucosal colorectal carcinoma from highgrade dysplasia and superficial submucosal invasive carcinoma by using high-definition colonoscopes without magnification.
Image library We assessed the component criteria and the overall classification by using an existing high-definition still image library of consecutive colorectal tumors, which was made up of single images of 22 tubular adenomas with low-grade dysplasia, 23 adenomatous lesions with highgrade dysplasia or superficially invasive (!1000 mm) submucosal carcinoma, and 35 deeply invasive (R1000 mm) submucosal carcinomas (a total of 80 images). The mean ( standard deviation) size of the lesions was 21.2 10.3 mm (range 6-60 mm). The images were captured with NBI by using high-definition colonoscopes without optical magnification (CF-H260AZI; Olympus Optical Co, Ltd, Tokyo, Japan) into a digital image filing system. One pathologist, who did not know the endoscopic diagnosis, assessed the histology according to the WHO classification.38
Phase 1: Evaluation of accuracy and reliability of histologic prediction by experienced colonoscopists In phase 1, we assessed the capacity of gastroenterologists with expert experience in NBI for polyp diagnosis (n Z 5: Europe [T.P., B.P.S.], Japan [S.T., Y.S.], and the United States [D.K.R., R.S.] to predict submucosal invasion. All endoscopists viewed polyp images on a high-definition monitor and scored the polyps as type 2 (comprising tubular adenoma with or without high-grade dysplasia and superficial submucosal carcinoma) or type 3 (SM-d carcinoma) and assigned a level of confidence to the prediction (high or low).26,37 Endoscopists were told to assign a prediction with high confidence when they had R90% certainty of the diagnosis.
Phase 2: Expansion of the classification and development of component criteria After confirmation in phase 1 that experienced endoscopists could accurately and reliably predict SM-d, a panel of experienced endoscopists (the authors of this study) further developed the NICE classification. We expanded the classification to include an additional category for lesions (type 3) with SM-d, which could be diagnosed without optical magnification by endoscopists without extensive experience in endoscopic imaging, chromoendoscopy, or pit-pattern diagnosis. We used a modified Delphi method39 to achieve consensus. During two working group meetings, the panel www.giejournal.org
Hayashi et al
Validation of the NICE classification
Figure 1. The NBI international colorectal endoscopic (NICE) classification. NBI, narrow-band imaging.
reviewed the results of phase 1 and developed criteria (color, vessels, and surface pattern) and descriptors for distinguishing SM-d carcinoma from other adenomatous lesions. Structured discussion and voting was again used to ensure equal participation and achieve consensus.
Phase 3: Validation of the criteria In phase 3, we evaluated the validity of the individual criteria for types 2 and 3 by using novice raters with no previous NBI experience, as we had done previously.37 Five medical students scored each colorectal tumor image for color, vessels, and surface pattern to assess the performance characteristics of the individual NICE criteria for types 2 and 3. To make sure that we evaluated the validity of the criteria in isolation, the students had no previous endoscopy or NBI experience, were given no training about colon tumors or NBI, and were not informed of the overall classification, of the relationship between the different criteria, or of the histology of the tumors they reviewed. The area of each image containing the tumor was marked, and students were shown exemplar images to demonstrate the typical appearances for each criterion (Fig. 2, although students were not shown the full classification). The score sheet presented www.giejournal.org
the descriptors in random order to make sure that students remained blinded to any patterns in the criteria.
Phase 4: Validation of the classification In the final phase, we assessed the performance characteristics and reliability of the overall classification by using raters trained in NBI diagnosis. We used the same students from phase 3, but after they completed a training session on colorectal tumors, NBI, and endoscopic diagnosis of submucosal invasion by using the NICE classification. During training, students were shown multiple exemplar still images of colorectal tumors over 30 minutes with expert commentary and explanation of NBI appearances. Students then repeated the assessment of the same set of colorectal tumor images by giving an overall prediction of the histology and assigning a level of confidence to their predictions.26,37
Statistical methods As in our previous study,37 the primary outcome measure for the predictive validity of the criteria and the classification was the negative predictive value of SM-d carcinoma. We calculated the accuracy, sensitivity, specificity, and negative and positive predictive values for each component of the classification and for the overall Volume 78, No. 4 : 2013 GASTROINTESTINAL ENDOSCOPY 627
Validation of the NICE classification
Hayashi et al
TABLE 1. Combined performance characteristics of expert predictions of deep submucosal invasion in still colorectal tumor images (n [ 6) Overall predictions, % (95% CI)
High-confidence predictions, % (95% CI)
Accuracy
75.6 (0.71-0.80)
84.1 (0.79-0.88)
Sensitivity
83.3 (0.76-0.89)
86.8 (0.79-0.92)
Specificity
70.0 (0.63-0.76)
81.5 (0.74-0.88)
NPV
85.4 (0.79-0.90)
86.9 (0.80-0.92)
PPV
66.5 (0.59-0.73)
81.4 (0.74-0.88)
CI, Confidence interval; NPV, negative predictive value; PPV, positive predictive value.
Phase 2: Development of the classification Figure 2. Study design. NBI, narrow-band imaging.
prediction by using the classification. We also calculated kappa statistics to assess for interrater reliability. The criterion standard for validation of predictions was the colorectal tumor histology. We calculated the sample size for the standardized image library used in phases 1, 2, and 4 by using the assumptions that 80% of predictions will be made with high confidence,26 and that the real accuracy was 90%; 80 tumors allowed the one-sided 95% confidence limit for measurement of performance characteristics to extend to 82%. Sufficient power for validity was achievable with a single rater; however, we chose to study 5 raters to ensure that the findings were reliable, acknowledging that the selection of the specific number of students (n Z 5) was relatively arbitrary.
The final version of the NICE classification, including types 1, 2, and 3, is shown in Figure 1.
Phase 3: Validation of the component criteria In phase 3, untrained novices (medical students) rated each polyp for each criterion. The performance characteristics of the criteria individually and in various combinations are shown in Table 3. The vessels criterion yielded the highest sensitivity and negative predictive value for an SM-d diagnosis, and surface pattern achieved the highest specificity. Color was the least sensitive and least predictive criterion, achieving lower accuracy than that of both vessels and surface pattern. When any 1 of the 3 SM-d carcinoma criteria were present, the sensitivity was 94.9%, and the negative predictive value was 95.9%, achieving higher sensitivity, specificity, and predictive value than did any of the 3 criteria alone.
Phase 4: Validation of the classification RESULTS The performance characteristics for ratings by colonoscopists experienced in NBI diagnosis are shown combined in Table 1 and individually in Table 2. Levels of sensitivity and negative predictive value for predicting SM-d with confidence were high, with negative predictive value reaching 90% in 3 of the raters (Table 2). Overall, 70% of predictions were made with high confidence, although this differed significantly between experts (Table 2: 53%-85%; P ! .001). The weighted multiple-rater kappa coefficient for recognition of SM-d was 0.66, and when limited to tumors rated with high confidence it was 0.97, indicating almost perfect agreement.
For phase 4, medical students were trained in NBI diagnosis of SM-d carcinoma and asked to provide a global prediction by using the component criteria and the overall classification. Compared with the before-training performance, the mean accuracies of the individual criteria among the 5 students’ rates after training improved to 80.3% for color (P Z .00628; kappa Z 0.139), 84.3% for vessels (P ! .0001; kappa Z 0.406), 85.0% for surface pattern (P ! .0001; kappa Z 0.458), and 84.3% overall (overall: P ! .0001; kappa Z 0.458). The combined sensitivity of an overall prediction with high confidence was 92%, and the negative predictive value was 92% (Table 4). The kappa value for the overall prediction of SM-d was 0.70 (95% confidence interval, 0.63-0.77), indicating substantial interobserver agreement.
628 GASTROINTESTINAL ENDOSCOPY Volume 78, No. 4 : 2013
www.giejournal.org
Phase 1: Experienced raters
Hayashi et al
Validation of the NICE classification
TABLE 2. Performance characteristics of individual expert raters for predictions made with high confidence in still colorectal polyp images % High-confidence predictions
Accuracy, % (95% CI)
Sensitivity, % (95% CI)
Specificity, % (95% CI)
NPV, % (95% CI)
PPV, % (95% CI)
1
75
88.9 (76.6-94.8)
90.0 (76.1-96.6)
88.0 (76.9-93.3)
91.7 (80.1-97.2)
85.7 (72.5-92.0)
2
85
88.2 (77.4-91.4)
78.2 (66.2-81.8)
96.4 (86.5-99.3)
84.3 (75.7-86.9)
94.7 (80.2-99.0)
3
53
87.5 (72.3-92.6)
83.3 (69.8-87.8)
92.9 (75.5-98.6)
93.8 (78.6-98.8)
81.3 (66.1-86.3)
4
72
74.4 (63.1-74.4)
100 (87.8-100)
52.2 (41.6-52.2)
100 (79.7-100)
64.5 (56.7-64.5)
5
72
81.4 (67.9-87.9)
90.0 (75.4-97.0)
73.9 (61.3-80.0)
89.5 (74.1-96.8)
75.0 (62.9-80.8)
6
62
83.8 (69.8-91.4)
80.0 (67.9-87.4)
88.2 (72.1-96.3)
78.9 (64.5-86.1)
88.9 (75.0-96.7)
Endoscopist
NPV, Negative predictive value; PPV, positive predictive value; CI, confidence interval.
TABLE 3. Performance characteristics of the NICE criteria* for the prediction of deep submucosal invasive carcinoma when applied by novice untrained students (phase 3, n [ 5) Accuracy, % (95% CI) Sensitivity, % (95% CI) Specificity, % (95% CI) PPV, % (95% CI) NPV, % (95% CI) Color
71.3 (66.7-75.4)
71.4 (66.2-76.2)
71.1 (67.0-74.8)
65.8 (61.0-70.2)
76.2 (71.8-80.2)
Vessels
78.7 (74.8-81.9)
88.0 (83.5-91.6)
71.6 (68.1-74.3)
70.6 (67.0-73.5)
88.5 (84.1-91.9)
Surface
77.2 (72.9-81.1)
72.6 (67.6-76.9)
80.9 (77.0-84.3)
74.7 (69.6-79.2)
79.1 (75.4-82.5)
Color or vessels
91.5 (88.6-93.4)
94.9 (91.5-97.1)
88.9 (86.3-90.6)
86.9 (83.8-88.9)
95.7 (92.9-97.6)
Color or surface
89.0 (85.6-91.7)
86.9 (82.9-89.9)
90.7 (87.6-93.1)
87.9 (83.9-91.0)
89.9 (86.8-92.2)
Vessels or surface
88.0 (84.5-90.7)
90.3 (86.3-93.3)
86.2 (83.1-88.6)
83.6 (79.9-86.4)
91.9 (88.7-94.5)
Color or vessels or surface
94.0 (91.3-95.8)
94.9 (91.7-97.0)
93.3 (90.9-95.0)
91.7 (88.7-93.7)
95.9 (93.4-97.6)
Color and vessels
58.5 (53.7-63.1)
64.6 (59.1-69.8)
53.8 (49.5-57.9)
52.1 (47.6-56.3)
66.1 (60.9-71.1)
Color and surface
59.5 (54.7-64.2)
57.1 (51.6-62.5)
61.3 (57.0-65.5)
53.5 (48.3-58.5)
64.8 (60.2-69.2)
Vessels and surface
68.0 (63.3-72.3)
70.3 (64.9-75.2)
66.2 (62.1-70.0)
61.8 (57.1-66.1)
74.1 (69.5-78.4)
Color and vessels and surface
51.5 (46.7-56.3)
54.9 (49.3-60.3)
48.9 (44.6-53.2)
45.5 (40.9-50.0)
58.2 (53.1-63.3)
NICE, NBI international colorectal endoscopic classification; NBI, narrow-band imaging; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value. *Using type 2 and type 3 of the NICE classification.
DISCUSSION The new concepts of optical diagnosis are based on the same principles as that of pathologic diagnosis; it is a diagnosis at the tissue level that is based on objective validated criteria. Objectivity is important. It is a departure from the prior commonly made endoscopic diagnosis, which is primarily based on subjectivity. The ability to make optical diagnosis that is based on objective validated criteria will allow endoscopic optical diagnosis to be viewed at the same level of diagnostic performance to that of pathologic diagnosis. Our efforts to validate the NICE classification
were specifically aimed to build the foundation of optical diagnosis of colorectal lesions. In this study, we show the validity of the expanded NICE classification and its components for differentiating deeply invasive (type 3) submucosal carcinoma from type 2dadenomadthat shows superficial submucosal invasion. Our results complement our prior validation study of the NICE classification (types 1 and 2) for the real time differentiation of hyperplastic and adenomatous colorectal polyps.37 In both studies, we first assembled an international collaboration to develop the classification and establish their expertise in endoscopic diagnosis. We
www.giejournal.org
Volume 78, No. 4 : 2013 GASTROINTESTINAL ENDOSCOPY 629
Validation of the NICE classification
Hayashi et al
TABLE 4. Performance characteristics of high-confidence, global predictions of deep submucosal invasive carcinoma by using the NICE classification* (phase 4) No. of high-confidence predictions (%)
Accuracy, % (95% CI)
Sensitivity, % (95% CI)
Specificity, % (95% CI)
PPV, % (95% CI)
NPV, % (95% CI)
1
18 (23.8)
100 (82.6-100)
100 (77.6-100)
100 (85.8-100)
100 (77.6-100)
100 (85.8-100)
2
38 (47.5)
84.2 (69.6-92.6)
82.4 (66.1-91.7)
85.7 (72.5-93.3)
82.4 (66.1-91.7)
85.7 (72.5-93.3)
3
17 (21.3)
94.1 (77.1-94.1)
92.9 (82.5-92.9)
100 (51.7-100)
100 (88.9-100)
75.0 (38.8-75.0)
4
61 (76.3)
93.4 (84.6-96.1)
91.2 (83.2-93.6)
96.3 (86.3-99.3)
96.9 (88.4-99.4)
89.7 (80.3-92.5)
5
66 (82.5)
86.4 (76.1-92.6)
82.8 (71.1-89.8)
89.2 (80.1-94.7)
85.7 (73.7-93.0)
86.8 (78.0-92.2)
90.0 (85.1-93.3)
91.8 (86.7-95.2)
88.3 (83.6-91.6)
88.1 (83.3-91.4)
91.9 (87.0-95.3)
Student
All
200 (50)
NICE, NBI international colorectal endoscopic classification; NBI, narrow-band imaging; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value. *Type 2 and type 3 of the NICE classification.
Figure 3. A, An example of the application of NICE classification to optically diagnose a depressed, 17-mm, colon neoplasm. B, The lesion was imaged by using NBI. The color of the lesion was brown relative to the background, with patchy white areas. C, Detailed viewing by using the NBI. There are areas of missing vessels. There are also areas missing a surface pattern. D, The lesion was well-differentiated adenocarcinoma with deep submucosal invasion and lymphatic involvement. NBI, narrow-band imaging (H&E, orig. mag. X 80).
then systematically assessed the new component criteria by using novice participants before demonstrating that these novices could be trained and apply the overall classification for SM-d carcinoma diagnosis with high levels of performance. The real-time endoscopic distinction between adenomas or superficial invasive submucosal carcinoma and
SM-d carcinoma is important in clinical practice to determine the appropriate therapeutic strategy. The validated classification will allow us to diagnose adenomas or superficial invasive submucosal carcinoma lesions, which can be curatively treated by endoscopic removal, from the deeply invasive carcinomas, which will require surgery (Figs. 3 and 4). In practice, endoscopists can perform
630 GASTROINTESTINAL ENDOSCOPY Volume 78, No. 4 : 2013
www.giejournal.org
Hayashi et al
Validation of the NICE classification
Figure 4. A, Another example of the application of NICE classification to optically diagnose a 1-cm sessile neoplasm. B, Optical diagnosis by using NBI was performed. C, Detailed imaging showed findings of NICE classification type 3. D, Pathology specimen after surgery showed welldifferentiated adenocarcinoma involving the deeper part of the submucosa. There was no lymphovascular involvement (H&E, orig. mag. X 80). NBI, narrow-band imaging.
resections that are likely curative and protect patients from the unnecessary risks of bleeding, perforation, and metastases. They can refer patients with lesions deemed to contain SM-d cancers to surgery, after marking the site and confirming the diagnosis pathologically. In Japan, by using optical magnification, NBI has been shown in multiple studies to enable accurate real-time determination of colorectal tumor histology.7,18 However, the global adoption of real-time diagnosis in routine clinical practice will require simple, standardized NBI criteria that can be easily learned and applied by both experienced and inexperienced endoscopistsdwithout optical magnification. The NICE classification was developed with the guiding principles of being simple and easy to use. To further ensure generalizability of the classification, we conducted this study by using images from colonoscopes without optical magnification. This study showed that careful observation of colorectal tumors by using highdefinition colonoscopes and NBI allows the tumor surface pattern to be well-visualized and its pathology optically diagnosed, all without optical magnification. This study has some limitations. Because adenomas with high-grade dysplasia, superficial submucosal invasive cancer, or deep submucosal cancer are uncommon, we used still images rather than prerecorded high-definition video to conduct the validation study. Second, we used
the same participants for phases 3 and 4 and have not yet prospectively assessed the application of the classification by practicing endoscopists during real-time clinical practice. The NICE classification itself is not without limitations. Currently, it does not specifically identify the sessile serrated adenoma and/or polyp.40 We have been unable to create a classification incorporating sessile serrated adenomas and/or polyps because there is currently low concordance of its diagnosis among pathologists.37 Without a pathologic criterion standard, it is difficult to include the serrated adenomatous polyps into the NICE classification. In conclusion, the NICE classification is a valid tool for real-time NBI diagnosis of SM-d carcinoma in images without high levels of optical magnification. Further real-time clinical studies are now required to confirm the validity and reliability of the NICE classification system for diagnosis of SM-d in clinical practice.
www.giejournal.org
Volume 78, No. 4 : 2013 GASTROINTESTINAL ENDOSCOPY 631
REFERENCES 1. Nakadoi T, Tanaka S, Kanao H, et al. Management of T1 colorectal carcinoma with special reference to criteria for curative endoscopic resection. J Gastroenterol Hepatol 2012;27:1057-62. 2. Watanabe T, Itabashi M, Shimada Y, et al. Japanese Society for Cancer of the Colon and Rectum (JSCCR) guidelines 2010 for the treatment of colorectal cancer. Int J Clin Oncol 2012;17:1-29.
Validation of the NICE classification 3. Meining A, von Delius S, Eames TM, et al. Risk factors for unfavorable outcomes after endoscopic removal of submucosal invasive colorectal tumors. Clin Gastroenterol Hepatol 2011;9:590-4. 4. Con SA, Saito Y, Matsuda T, et al. Application of endoscopic submucosal dissection for removal of deep invasive submucosal colon carcinoma. Gastrointest Endosc 2010;72:1217-25. 5. Tanaka S, Sano Y. Aim to unify the narrow band imaging (NBI) magnifying classification for colorectal tumors: current status in Japan from a summary of the consensus symposium in the 79th Annual Meeting of the Japan Gastroenterological Endoscopy Society. Dig Endosc 2011;(23 suppl 1):131-9. 6. Oka S, Tanaka S, Takata S, et al. Clinical usefulness of narrow band imaging magnifying classification for colorectal tumors based on both surface pattern and microvessel features. Dig Endosc 2011;(23 suppl 1):101-5. 7. Oba S, Tanaka S, Sano Y, et al. Current status of narrow-band imaging magnifying colonoscopy for colorectal neoplasia in Japan. Digestion 2011;83:167-72. 8. Oba S, Tanaka S, Oka S, et al. Characterization of colorectal tumors using narrow-band imaging magnification: combined diagnosis with both pit pattern and microvessel features. Scand J Gastroenterol 2010;45:1084-92. 9. Kanao H, Tanaka S, Oka S, et al. Narrow-band imaging magnification predicts the histology and invasion depth of colorectal tumors. Gastrointest Endosc 2009;69:631-6. 10. Hirata M, Tanaka S, Oka S, et al. Magnifying endoscopy with narrow band imaging for diagnosis of colorectal tumors. Gastrointest Endosc 2007;65:988-95. 11. Hirata M, Tanaka S, Oka S, et al. Evaluation of microvessels in colorectal tumors by narrow band imaging magnification. Gastrointest Endosc 2007;66:945-52. 12. Ikematsu H, Saito Y, Yamano H. Comparative evaluation of endoscopic factors from conventional colonoscopy and narrow-band imaging of colorectal lesions. Dig Endosc 2011;(23 suppl 1):95-100. 13. Ikematsu H, Matsuda T, Emura F, et al. Efficacy of capillary pattern type IIIA/IIIB by magnifying narrow band imaging for estimating depth of invasion of early colorectal neoplasms. BMC Gastroenterol 2010;10:33. 14. Sano Y, Ikematsu H, Fu KI, et al. Meshed capillary vessels by use of narrow-band imaging for differential diagnosis of small colorectal polyps. Gastrointest Endosc 2009;69:278-83. 15. Sano Y, Ikematsu H, Fu KI, et al. Meshed capillary vessels by use of narrow-band imaging for differential diagnosis of neoplastic and nonneoplastic colonic polyps. Am J Gastroenterol 2006;101:2711-6. 16. Katagiri A, Fu KI, Sano Y, et al. Narrow band imaging with magnifying colonoscopy as diagnostic tool for predicting histology of early colorectal neoplasia. Aliment Pharmacol Ther 2008;27:1269-74. 17. Machida H, Sano Y, Hamamoto Y, et al. Narrow-band imaging in the diagnosis of colorectal mucosal lesions: a pilot study. Endoscopy 2004;36:1094-8. 18. Wada Y, Kudo SE, Kashida H, et al. Diagnosis of colorectal lesions with the magnifying narrow-band imaging system. Gastrointest Endosc 2009;70:522-31. 19. East JE, Suzuki N, Saunders BP. Comparison of magnified pit pattern interpretation with narrow band imaging versus chromoendoscopy for diminutive colonic polyps: a pilot study. Gastrointest Endosc 2007; 66:310-6. 20. My Su, Hsu CM, Ho YP, et al. Comparative study of conventional colonoscopy, chromoendoscopy, and narrow-band imaging systems in differential diagnosis of neoplastic and nonneoplastic colonic polyps. Am J Gastroenterol 2006;101:2711-6. 21. East JE, Suzuki N, Bassett P, et al. Narrow band imaging with magnification for the characterization of small and diminutive colonic polyps: pit pattern and vascular pattern intensity. Endoscopy 2008;40:811-7. 22. Chiu HM, Chang CY, Chen CC, et al. A prospective comparative study of narrow-band imaging, chromoendoscopy, and conventional colonoscopy in the diagnosis of colorectal neoplasia. Gut 2007;39:1092-6.
632 GASTROINTESTINAL ENDOSCOPY Volume 78, No. 4 : 2013
Hayashi et al 23. Tischendorf JJ, Wasmuth HE, Koch A, et al. Value of magnifying chromoendoscopy and narrow band imaging (NBI) in classifying colorectal polyps: a prospective controlled study. Endoscopy 2007;39:1092-6. 24. Chang CC, Hsieh CR, Lou HY, et al. Comparative study of conventional colonoscopy, magnifying chromoendoscopy, and magnifying narrow-band imaging system in the differential diagnosis of small colonic polyps between trainee and experienced endoscopist. Int J Colorectal Dis 2009;24:1413-9. 25. Singh R, Nordeen N, Mei SL, et al. West meets East: preliminary results of narrow band imaging with optical magnification in the diagnosis of colorectal lesions: a multicenter Australian study using the modified Sano’s classification. Dig Endosc 2011;(23 Suppl 1):126-30. 26. Rex DK. Narrow-band imaging without optical magnification for histologic analysis of colorectal polyps. Gastroenterology 2009;136: 1174-81. 27. Rex DK, Helbig CC. High yields of small and flat adenomas with highdefinition colonoscopes using either white light or narrow band imaging. Gastroenterology 2007;133:42-7. 28. Rastogi A, Bansal A, Wani S, et al. Narrow-band imaging colonoscopyd a pilot feasibility study for the detection of polyps and correlation of surface patterns with polyp histologic diagnosis. Gastrointest Endosc 2008;67:280-6. 29. Rastogi A, Pondugula K, Bansal A, et al. Recognition of surface mucosal and vascular patterns of colon polyps by using narrow-band imaging: interobserver and intraobserver agreement and prediction of polyp histology. Gastrointest Endosc 2009;69:716-22. 30. Rastogi A, Keighley J, Singh V, et al. High accuracy of narrow band imaging without magnification for the real-time characterization of polyp histology and its comparison with high-definition white light colonoscopy: a prospective study. Am J Gastroenterol 2009;104: 2422-30. 31. Rogart JN, Jain D, Siddiqui UD, et al. Narrow-band imaging without high magnification to differentiate polyps during real-time colonoscopy: improvement with experience. Gastrointest Endosc 2008;68: 1136-45. 32. Sikka S, Ringold DA, Jonnalagadda S, et al. Comparison of white light and narrow band high definition images in predicting colon polyp histology, using standard colonoscopes without optical magnification. Endoscopy 2008;40:818-22. 33. Ignjatovic A, East JE, Guenther T, et al. What is the most reliable imaging modality for small colonic polyp characterization? Study of white-light, autofluorescence, and narrow-band imaging. Endoscopy 2011;43:94-9. 34. Zhon QJ, Yang JM, Fei BY, et al. Narrow-band imaging endoscopy with and without magnification in diagnosis of colorectal neoplasia. World J Gastroenterol 2011;17:666-70. 35. Lee CK, Lee SH, Hwangbo Y. Narrow-band imaging versus I-Scan for the real-time histological prediction of diminutive colonic polyps: a prospective comparative study by using the simple unified endoscopic classification. Gastrointest Endosc 2011;74:603-9. 36. Rastogi A, Early DS, Gupta N, et al. Randomized, controlled trial of standard-definition white-light, high-definition white- light, and narrow-band imaging colonoscopy for the detection of colon polyps and prediction of polyp histology. Gastrointest Endosc 2011;74: 593-602. 37. Hewett DG, Kaltenbach T, Sano Y, et al. Validation of a simple classification system for endoscopic diagnosis of small colorectal polyps using narrow-band imaging. Gastroenterology 2012;143: 599-607. 38. Bosman FT, Carneiro F, Hruban RH, et al, editors. WHO classification of tumours of the digestive system. Lyon (France): IARC Press; 2010. p. 160-5. 39. Milholland AV, Wheeler SG, Heieck JJ. Medical assessment by a Delphi group opinion technic. N Engl J Med 1973;288:1272-5. 40. Rex DK, Ahnen DJ, Baron JA, et al. Serrated lesions of the colorectum: review and recommendations from an expert panel. Am J Gastroenterol 2012;107:1315-29.
www.giejournal.org
Endoscopic prediction of deep submucosal invasive carcinoma: validation of the Narrow-Band Imaging International Colorectal Endoscopic (NICE) classification Nana Hayashi, MD,1 Shinji Tanaka, MD, PhD,1 David G. Hewett, MD, PhD,2 Tonya R. Kaltenbach, MD, PhD,3 Yasushi Sano, MD, PhD,4 Thierry Ponchon, MD, PhD,5 Brian P. Saunders, MD, PhD,6 Douglas K. Rex, MD, PhD,7 Roy M. Soetikno, MD, PhD3 Hiroshima, Japan
Background: A simple endoscopic classification to accurately predict deep submucosal invasive (SM-d) carcinoma would be clinically useful. Objective: To develop and assess the validity of the NBI international colorectal endoscopic (NICE) classification for the characterization of SM-d carcinoma. Design: The study was conducted in 4 phases: (1) evaluation of endoscopic differentiation by NBI-experienced colonoscopists; (2) extension of the NICE classification to incorporate SM-d (type 3) by using a modified Delphi method; (3) prospective validation of the individual criteria by inexperienced participants, by using highdefinition still images without magnification of known histology; and (4) prospective validation of the individual criteria and overall classification by inexperienced participants after training. Setting: Japanese academic unit. Main Outcome Measurements: Performance characteristics of the NICE criteria (phase 3) and overall classification (phase 4) for SM-d carcinoma; sensitivity, specificity, predictive values, and accuracy. Results: We expanded the NICE classification for the endoscopic diagnosis of SM-d carcinoma (type 3) and established the predictive validity of its individual components. The negative predictive values of the individual criteria for diagnosis of SM-d carcinoma were 76.2% (color), 88.5% (vessels), and 79.1% (surface pattern). When any 1 of the 3 SM-d criteria was present, the sensitivity was 94.9%, and the negative predictive value was 95.9%. The overall sensitivity and negative predictive value of a global, high-confidence prediction of SM-d carcinoma was 92%. Interobserver agreement for an overall SM-d carcinoma prediction was substantial (kappa 0.70). Limitations: Single Japanese center, use of still images without prospective clinical evaluation. Conclusion: The NICE classification is a valid tool for predicting SM-d carcinomas in colorectal tumors. (Gastrointest Endosc 2013;78:625-32.)
Abbreviations: NBI, narrow-band imaging; NICE, NBI international colorectal endoscopic; SM-d, deep submucosal invasion. DISCLOSURE: All authors disclosed no financial relationships relevant to this publication. See CME section; p. 637.
Use your mobile device to scan this QR code and watch the author interview. Download a free QR code scanner by searching “QR Scanner” in your mobile device’s app store. www.giejournal.org
Copyright ª 2013 by the American Society for Gastrointestinal Endoscopy 0016-5107/$36.00 http://dx.doi.org/10.1016/j.gie.2013.04.185 Received October 9, 2012. Accepted April 20, 2013. Current affiliations: Department of Endoscopy, Hiroshima University Hospital, Hiroshima, Japan (1), University of Queensland School of Medicine, Australia (2), Veterans Affairs Palo Alto Health Care System, Palo Alto, California (3), Sano Hospital, Hyogo, Japan (4), Hôpital Edouard-Herriot, Lyon, France (5), St. Mark’s Hospital, London, United Kingdom (6), Indiana University, Indianapolis, Indiana, USA (7). Reprint requests: Shinji Tanaka, MD, PhD, Department of Endoscopy, Hiroshima University, 1-2-3 Kasumi, Minami-ku, Hiroshima 734-8551, Japan.
Volume 78, No. 4 : 2013 GASTROINTESTINAL ENDOSCOPY 625
Validation of the NICE classification
The potential for endoscopic treatment of early colorectal cancer is being realized through recent advances in endoscopic imaging and resection techniques. These resections require the ability to select the appropriate neoplastic lesions because patients with high-grade dysplasia and submucosal invasive carcinoma with an invasion depth of !1000 mm (superficial submucosal invasion), absence of tumor budding, and absence of lymphovascular invasion are likely to be cured after endoscopic resection,1-3 whereas those with deeper submucosal invasion have up to a 10% risk of lymph node metastasis.1,2 The endoscopic resection of deep (R1000 mm) submucosal invasive (SM-d) carcinoma, also is associated with an increased risk of bleeding and perforation.4 Clinical experience with detailed inspection of the continuous spectrum of adenomas as they progress to carcinoma suggests that their appearance under narrow-band imaging (NBI) can be artificially classified into 2 typesd type 2 and type 3.5 Type 3 includes SM-d carcinoma, whereas type 2 includes adenomas to superficial invasive carcinoma. A number of studies have suggested the potential validity of these methods of classification,6-36 but the validity of their component criteria has not been established. In this study, we evaluated the predictive validity of an expanded classification for the endoscopic differentiation of deep submucosal colorectal carcinoma from tubular adenomas, adenomas with high-grade dysplasia, and superficial submucosal carcinoma. Specifically, we aimed to (1) expand the NBI international colorectal endoscopic (NICE) classification (Fig. 1) to include SM-d carcinoma (type 3), in addition to type 1 (hyperplastic) and type 2 (adenomatous to superficial submucosal colorectal carcinoma)5,7,37; (2) evaluate the predictive validity of the components of the classification; and (3) evaluate the validity and performance characteristics of the overall classification for distinguishing type 3 from types 1 and 2.
METHODS Study design We conducted a prospective validation study of the NICE classification for the diagnosis of SM-d carcinoma in colorectal tumors by using NBI without optical magnification. Our approach here reflects the one used in our previous validation study of classifying type 1 versus type 2.37 Specifically, the study was conducted in 4 phases (Fig. 2): (1) evaluation of the accuracy of NBI-experienced colonoscopists in predicting SM-d, (2) development of an expanded NICE classification to incorporate a further category (type 3) for SM-d, (3) validation of the component criteria by testing the criteria in endoscopy-naïve participants, and (4) assessment of the performance and reliability of the overall classification for predicting histology in participants trained in NBI diagnosis. The Institutional Review Board at Hiroshima University Hospital approved the study. 626 GASTROINTESTINAL ENDOSCOPY Volume 78, No. 4 : 2013
Hayashi et al
Take-home message The authors aimed to establish the predictive validity of simple classification for the endoscopic differentiation of deep submucosal colorectal carcinoma from highgrade dysplasia and superficial submucosal invasive carcinoma by using high-definition colonoscopes without magnification.
Image library We assessed the component criteria and the overall classification by using an existing high-definition still image library of consecutive colorectal tumors, which was made up of single images of 22 tubular adenomas with low-grade dysplasia, 23 adenomatous lesions with highgrade dysplasia or superficially invasive (!1000 mm) submucosal carcinoma, and 35 deeply invasive (R1000 mm) submucosal carcinomas (a total of 80 images). The mean ( standard deviation) size of the lesions was 21.2 10.3 mm (range 6-60 mm). The images were captured with NBI by using high-definition colonoscopes without optical magnification (CF-H260AZI; Olympus Optical Co, Ltd, Tokyo, Japan) into a digital image filing system. One pathologist, who did not know the endoscopic diagnosis, assessed the histology according to the WHO classification.38
Phase 1: Evaluation of accuracy and reliability of histologic prediction by experienced colonoscopists In phase 1, we assessed the capacity of gastroenterologists with expert experience in NBI for polyp diagnosis (n Z 5: Europe [T.P., B.P.S.], Japan [S.T., Y.S.], and the United States [D.K.R., R.S.] to predict submucosal invasion. All endoscopists viewed polyp images on a high-definition monitor and scored the polyps as type 2 (comprising tubular adenoma with or without high-grade dysplasia and superficial submucosal carcinoma) or type 3 (SM-d carcinoma) and assigned a level of confidence to the prediction (high or low).26,37 Endoscopists were told to assign a prediction with high confidence when they had R90% certainty of the diagnosis.
Phase 2: Expansion of the classification and development of component criteria After confirmation in phase 1 that experienced endoscopists could accurately and reliably predict SM-d, a panel of experienced endoscopists (the authors of this study) further developed the NICE classification. We expanded the classification to include an additional category for lesions (type 3) with SM-d, which could be diagnosed without optical magnification by endoscopists without extensive experience in endoscopic imaging, chromoendoscopy, or pit-pattern diagnosis. We used a modified Delphi method39 to achieve consensus. During two working group meetings, the panel www.giejournal.org
Hayashi et al
Validation of the NICE classification
Figure 1. The NBI international colorectal endoscopic (NICE) classification. NBI, narrow-band imaging.
reviewed the results of phase 1 and developed criteria (color, vessels, and surface pattern) and descriptors for distinguishing SM-d carcinoma from other adenomatous lesions. Structured discussion and voting was again used to ensure equal participation and achieve consensus.
Phase 3: Validation of the criteria In phase 3, we evaluated the validity of the individual criteria for types 2 and 3 by using novice raters with no previous NBI experience, as we had done previously.37 Five medical students scored each colorectal tumor image for color, vessels, and surface pattern to assess the performance characteristics of the individual NICE criteria for types 2 and 3. To make sure that we evaluated the validity of the criteria in isolation, the students had no previous endoscopy or NBI experience, were given no training about colon tumors or NBI, and were not informed of the overall classification, of the relationship between the different criteria, or of the histology of the tumors they reviewed. The area of each image containing the tumor was marked, and students were shown exemplar images to demonstrate the typical appearances for each criterion (Fig. 2, although students were not shown the full classification). The score sheet presented www.giejournal.org
the descriptors in random order to make sure that students remained blinded to any patterns in the criteria.
Phase 4: Validation of the classification In the final phase, we assessed the performance characteristics and reliability of the overall classification by using raters trained in NBI diagnosis. We used the same students from phase 3, but after they completed a training session on colorectal tumors, NBI, and endoscopic diagnosis of submucosal invasion by using the NICE classification. During training, students were shown multiple exemplar still images of colorectal tumors over 30 minutes with expert commentary and explanation of NBI appearances. Students then repeated the assessment of the same set of colorectal tumor images by giving an overall prediction of the histology and assigning a level of confidence to their predictions.26,37
Statistical methods As in our previous study,37 the primary outcome measure for the predictive validity of the criteria and the classification was the negative predictive value of SM-d carcinoma. We calculated the accuracy, sensitivity, specificity, and negative and positive predictive values for each component of the classification and for the overall Volume 78, No. 4 : 2013 GASTROINTESTINAL ENDOSCOPY 627
Validation of the NICE classification
Hayashi et al
TABLE 1. Combined performance characteristics of expert predictions of deep submucosal invasion in still colorectal tumor images (n [ 6) Overall predictions, % (95% CI)
High-confidence predictions, % (95% CI)
Accuracy
75.6 (0.71-0.80)
84.1 (0.79-0.88)
Sensitivity
83.3 (0.76-0.89)
86.8 (0.79-0.92)
Specificity
70.0 (0.63-0.76)
81.5 (0.74-0.88)
NPV
85.4 (0.79-0.90)
86.9 (0.80-0.92)
PPV
66.5 (0.59-0.73)
81.4 (0.74-0.88)
CI, Confidence interval; NPV, negative predictive value; PPV, positive predictive value.
Phase 2: Development of the classification Figure 2. Study design. NBI, narrow-band imaging.
prediction by using the classification. We also calculated kappa statistics to assess for interrater reliability. The criterion standard for validation of predictions was the colorectal tumor histology. We calculated the sample size for the standardized image library used in phases 1, 2, and 4 by using the assumptions that 80% of predictions will be made with high confidence,26 and that the real accuracy was 90%; 80 tumors allowed the one-sided 95% confidence limit for measurement of performance characteristics to extend to 82%. Sufficient power for validity was achievable with a single rater; however, we chose to study 5 raters to ensure that the findings were reliable, acknowledging that the selection of the specific number of students (n Z 5) was relatively arbitrary.
The final version of the NICE classification, including types 1, 2, and 3, is shown in Figure 1.
Phase 3: Validation of the component criteria In phase 3, untrained novices (medical students) rated each polyp for each criterion. The performance characteristics of the criteria individually and in various combinations are shown in Table 3. The vessels criterion yielded the highest sensitivity and negative predictive value for an SM-d diagnosis, and surface pattern achieved the highest specificity. Color was the least sensitive and least predictive criterion, achieving lower accuracy than that of both vessels and surface pattern. When any 1 of the 3 SM-d carcinoma criteria were present, the sensitivity was 94.9%, and the negative predictive value was 95.9%, achieving higher sensitivity, specificity, and predictive value than did any of the 3 criteria alone.
Phase 4: Validation of the classification RESULTS The performance characteristics for ratings by colonoscopists experienced in NBI diagnosis are shown combined in Table 1 and individually in Table 2. Levels of sensitivity and negative predictive value for predicting SM-d with confidence were high, with negative predictive value reaching 90% in 3 of the raters (Table 2). Overall, 70% of predictions were made with high confidence, although this differed significantly between experts (Table 2: 53%-85%; P ! .001). The weighted multiple-rater kappa coefficient for recognition of SM-d was 0.66, and when limited to tumors rated with high confidence it was 0.97, indicating almost perfect agreement.
For phase 4, medical students were trained in NBI diagnosis of SM-d carcinoma and asked to provide a global prediction by using the component criteria and the overall classification. Compared with the before-training performance, the mean accuracies of the individual criteria among the 5 students’ rates after training improved to 80.3% for color (P Z .00628; kappa Z 0.139), 84.3% for vessels (P ! .0001; kappa Z 0.406), 85.0% for surface pattern (P ! .0001; kappa Z 0.458), and 84.3% overall (overall: P ! .0001; kappa Z 0.458). The combined sensitivity of an overall prediction with high confidence was 92%, and the negative predictive value was 92% (Table 4). The kappa value for the overall prediction of SM-d was 0.70 (95% confidence interval, 0.63-0.77), indicating substantial interobserver agreement.
628 GASTROINTESTINAL ENDOSCOPY Volume 78, No. 4 : 2013
www.giejournal.org
Phase 1: Experienced raters
Hayashi et al
Validation of the NICE classification
TABLE 2. Performance characteristics of individual expert raters for predictions made with high confidence in still colorectal polyp images % High-confidence predictions
Accuracy, % (95% CI)
Sensitivity, % (95% CI)
Specificity, % (95% CI)
NPV, % (95% CI)
PPV, % (95% CI)
1
75
88.9 (76.6-94.8)
90.0 (76.1-96.6)
88.0 (76.9-93.3)
91.7 (80.1-97.2)
85.7 (72.5-92.0)
2
85
88.2 (77.4-91.4)
78.2 (66.2-81.8)
96.4 (86.5-99.3)
84.3 (75.7-86.9)
94.7 (80.2-99.0)
3
53
87.5 (72.3-92.6)
83.3 (69.8-87.8)
92.9 (75.5-98.6)
93.8 (78.6-98.8)
81.3 (66.1-86.3)
4
72
74.4 (63.1-74.4)
100 (87.8-100)
52.2 (41.6-52.2)
100 (79.7-100)
64.5 (56.7-64.5)
5
72
81.4 (67.9-87.9)
90.0 (75.4-97.0)
73.9 (61.3-80.0)
89.5 (74.1-96.8)
75.0 (62.9-80.8)
6
62
83.8 (69.8-91.4)
80.0 (67.9-87.4)
88.2 (72.1-96.3)
78.9 (64.5-86.1)
88.9 (75.0-96.7)
Endoscopist
NPV, Negative predictive value; PPV, positive predictive value; CI, confidence interval.
TABLE 3. Performance characteristics of the NICE criteria* for the prediction of deep submucosal invasive carcinoma when applied by novice untrained students (phase 3, n [ 5) Accuracy, % (95% CI) Sensitivity, % (95% CI) Specificity, % (95% CI) PPV, % (95% CI) NPV, % (95% CI) Color
71.3 (66.7-75.4)
71.4 (66.2-76.2)
71.1 (67.0-74.8)
65.8 (61.0-70.2)
76.2 (71.8-80.2)
Vessels
78.7 (74.8-81.9)
88.0 (83.5-91.6)
71.6 (68.1-74.3)
70.6 (67.0-73.5)
88.5 (84.1-91.9)
Surface
77.2 (72.9-81.1)
72.6 (67.6-76.9)
80.9 (77.0-84.3)
74.7 (69.6-79.2)
79.1 (75.4-82.5)
Color or vessels
91.5 (88.6-93.4)
94.9 (91.5-97.1)
88.9 (86.3-90.6)
86.9 (83.8-88.9)
95.7 (92.9-97.6)
Color or surface
89.0 (85.6-91.7)
86.9 (82.9-89.9)
90.7 (87.6-93.1)
87.9 (83.9-91.0)
89.9 (86.8-92.2)
Vessels or surface
88.0 (84.5-90.7)
90.3 (86.3-93.3)
86.2 (83.1-88.6)
83.6 (79.9-86.4)
91.9 (88.7-94.5)
Color or vessels or surface
94.0 (91.3-95.8)
94.9 (91.7-97.0)
93.3 (90.9-95.0)
91.7 (88.7-93.7)
95.9 (93.4-97.6)
Color and vessels
58.5 (53.7-63.1)
64.6 (59.1-69.8)
53.8 (49.5-57.9)
52.1 (47.6-56.3)
66.1 (60.9-71.1)
Color and surface
59.5 (54.7-64.2)
57.1 (51.6-62.5)
61.3 (57.0-65.5)
53.5 (48.3-58.5)
64.8 (60.2-69.2)
Vessels and surface
68.0 (63.3-72.3)
70.3 (64.9-75.2)
66.2 (62.1-70.0)
61.8 (57.1-66.1)
74.1 (69.5-78.4)
Color and vessels and surface
51.5 (46.7-56.3)
54.9 (49.3-60.3)
48.9 (44.6-53.2)
45.5 (40.9-50.0)
58.2 (53.1-63.3)
NICE, NBI international colorectal endoscopic classification; NBI, narrow-band imaging; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value. *Using type 2 and type 3 of the NICE classification.
DISCUSSION The new concepts of optical diagnosis are based on the same principles as that of pathologic diagnosis; it is a diagnosis at the tissue level that is based on objective validated criteria. Objectivity is important. It is a departure from the prior commonly made endoscopic diagnosis, which is primarily based on subjectivity. The ability to make optical diagnosis that is based on objective validated criteria will allow endoscopic optical diagnosis to be viewed at the same level of diagnostic performance to that of pathologic diagnosis. Our efforts to validate the NICE classification
were specifically aimed to build the foundation of optical diagnosis of colorectal lesions. In this study, we show the validity of the expanded NICE classification and its components for differentiating deeply invasive (type 3) submucosal carcinoma from type 2dadenomadthat shows superficial submucosal invasion. Our results complement our prior validation study of the NICE classification (types 1 and 2) for the real time differentiation of hyperplastic and adenomatous colorectal polyps.37 In both studies, we first assembled an international collaboration to develop the classification and establish their expertise in endoscopic diagnosis. We
www.giejournal.org
Volume 78, No. 4 : 2013 GASTROINTESTINAL ENDOSCOPY 629
Validation of the NICE classification
Hayashi et al
TABLE 4. Performance characteristics of high-confidence, global predictions of deep submucosal invasive carcinoma by using the NICE classification* (phase 4) No. of high-confidence predictions (%)
Accuracy, % (95% CI)
Sensitivity, % (95% CI)
Specificity, % (95% CI)
PPV, % (95% CI)
NPV, % (95% CI)
1
18 (23.8)
100 (82.6-100)
100 (77.6-100)
100 (85.8-100)
100 (77.6-100)
100 (85.8-100)
2
38 (47.5)
84.2 (69.6-92.6)
82.4 (66.1-91.7)
85.7 (72.5-93.3)
82.4 (66.1-91.7)
85.7 (72.5-93.3)
3
17 (21.3)
94.1 (77.1-94.1)
92.9 (82.5-92.9)
100 (51.7-100)
100 (88.9-100)
75.0 (38.8-75.0)
4
61 (76.3)
93.4 (84.6-96.1)
91.2 (83.2-93.6)
96.3 (86.3-99.3)
96.9 (88.4-99.4)
89.7 (80.3-92.5)
5
66 (82.5)
86.4 (76.1-92.6)
82.8 (71.1-89.8)
89.2 (80.1-94.7)
85.7 (73.7-93.0)
86.8 (78.0-92.2)
90.0 (85.1-93.3)
91.8 (86.7-95.2)
88.3 (83.6-91.6)
88.1 (83.3-91.4)
91.9 (87.0-95.3)
Student
All
200 (50)
NICE, NBI international colorectal endoscopic classification; NBI, narrow-band imaging; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value. *Type 2 and type 3 of the NICE classification.
Figure 3. A, An example of the application of NICE classification to optically diagnose a depressed, 17-mm, colon neoplasm. B, The lesion was imaged by using NBI. The color of the lesion was brown relative to the background, with patchy white areas. C, Detailed viewing by using the NBI. There are areas of missing vessels. There are also areas missing a surface pattern. D, The lesion was well-differentiated adenocarcinoma with deep submucosal invasion and lymphatic involvement. NBI, narrow-band imaging (H&E, orig. mag. X 80).
then systematically assessed the new component criteria by using novice participants before demonstrating that these novices could be trained and apply the overall classification for SM-d carcinoma diagnosis with high levels of performance. The real-time endoscopic distinction between adenomas or superficial invasive submucosal carcinoma and
SM-d carcinoma is important in clinical practice to determine the appropriate therapeutic strategy. The validated classification will allow us to diagnose adenomas or superficial invasive submucosal carcinoma lesions, which can be curatively treated by endoscopic removal, from the deeply invasive carcinomas, which will require surgery (Figs. 3 and 4). In practice, endoscopists can perform
630 GASTROINTESTINAL ENDOSCOPY Volume 78, No. 4 : 2013
www.giejournal.org
Hayashi et al
Validation of the NICE classification
Figure 4. A, Another example of the application of NICE classification to optically diagnose a 1-cm sessile neoplasm. B, Optical diagnosis by using NBI was performed. C, Detailed imaging showed findings of NICE classification type 3. D, Pathology specimen after surgery showed welldifferentiated adenocarcinoma involving the deeper part of the submucosa. There was no lymphovascular involvement (H&E, orig. mag. X 80). NBI, narrow-band imaging.
resections that are likely curative and protect patients from the unnecessary risks of bleeding, perforation, and metastases. They can refer patients with lesions deemed to contain SM-d cancers to surgery, after marking the site and confirming the diagnosis pathologically. In Japan, by using optical magnification, NBI has been shown in multiple studies to enable accurate real-time determination of colorectal tumor histology.7,18 However, the global adoption of real-time diagnosis in routine clinical practice will require simple, standardized NBI criteria that can be easily learned and applied by both experienced and inexperienced endoscopistsdwithout optical magnification. The NICE classification was developed with the guiding principles of being simple and easy to use. To further ensure generalizability of the classification, we conducted this study by using images from colonoscopes without optical magnification. This study showed that careful observation of colorectal tumors by using highdefinition colonoscopes and NBI allows the tumor surface pattern to be well-visualized and its pathology optically diagnosed, all without optical magnification. This study has some limitations. Because adenomas with high-grade dysplasia, superficial submucosal invasive cancer, or deep submucosal cancer are uncommon, we used still images rather than prerecorded high-definition video to conduct the validation study. Second, we used
the same participants for phases 3 and 4 and have not yet prospectively assessed the application of the classification by practicing endoscopists during real-time clinical practice. The NICE classification itself is not without limitations. Currently, it does not specifically identify the sessile serrated adenoma and/or polyp.40 We have been unable to create a classification incorporating sessile serrated adenomas and/or polyps because there is currently low concordance of its diagnosis among pathologists.37 Without a pathologic criterion standard, it is difficult to include the serrated adenomatous polyps into the NICE classification. In conclusion, the NICE classification is a valid tool for real-time NBI diagnosis of SM-d carcinoma in images without high levels of optical magnification. Further real-time clinical studies are now required to confirm the validity and reliability of the NICE classification system for diagnosis of SM-d in clinical practice.
www.giejournal.org
Volume 78, No. 4 : 2013 GASTROINTESTINAL ENDOSCOPY 631
REFERENCES 1. Nakadoi T, Tanaka S, Kanao H, et al. Management of T1 colorectal carcinoma with special reference to criteria for curative endoscopic resection. J Gastroenterol Hepatol 2012;27:1057-62. 2. Watanabe T, Itabashi M, Shimada Y, et al. Japanese Society for Cancer of the Colon and Rectum (JSCCR) guidelines 2010 for the treatment of colorectal cancer. Int J Clin Oncol 2012;17:1-29.
Validation of the NICE classification 3. Meining A, von Delius S, Eames TM, et al. Risk factors for unfavorable outcomes after endoscopic removal of submucosal invasive colorectal tumors. Clin Gastroenterol Hepatol 2011;9:590-4. 4. Con SA, Saito Y, Matsuda T, et al. Application of endoscopic submucosal dissection for removal of deep invasive submucosal colon carcinoma. Gastrointest Endosc 2010;72:1217-25. 5. Tanaka S, Sano Y. Aim to unify the narrow band imaging (NBI) magnifying classification for colorectal tumors: current status in Japan from a summary of the consensus symposium in the 79th Annual Meeting of the Japan Gastroenterological Endoscopy Society. Dig Endosc 2011;(23 suppl 1):131-9. 6. Oka S, Tanaka S, Takata S, et al. Clinical usefulness of narrow band imaging magnifying classification for colorectal tumors based on both surface pattern and microvessel features. Dig Endosc 2011;(23 suppl 1):101-5. 7. Oba S, Tanaka S, Sano Y, et al. Current status of narrow-band imaging magnifying colonoscopy for colorectal neoplasia in Japan. Digestion 2011;83:167-72. 8. Oba S, Tanaka S, Oka S, et al. Characterization of colorectal tumors using narrow-band imaging magnification: combined diagnosis with both pit pattern and microvessel features. Scand J Gastroenterol 2010;45:1084-92. 9. Kanao H, Tanaka S, Oka S, et al. Narrow-band imaging magnification predicts the histology and invasion depth of colorectal tumors. Gastrointest Endosc 2009;69:631-6. 10. Hirata M, Tanaka S, Oka S, et al. Magnifying endoscopy with narrow band imaging for diagnosis of colorectal tumors. Gastrointest Endosc 2007;65:988-95. 11. Hirata M, Tanaka S, Oka S, et al. Evaluation of microvessels in colorectal tumors by narrow band imaging magnification. Gastrointest Endosc 2007;66:945-52. 12. Ikematsu H, Saito Y, Yamano H. Comparative evaluation of endoscopic factors from conventional colonoscopy and narrow-band imaging of colorectal lesions. Dig Endosc 2011;(23 suppl 1):95-100. 13. Ikematsu H, Matsuda T, Emura F, et al. Efficacy of capillary pattern type IIIA/IIIB by magnifying narrow band imaging for estimating depth of invasion of early colorectal neoplasms. BMC Gastroenterol 2010;10:33. 14. Sano Y, Ikematsu H, Fu KI, et al. Meshed capillary vessels by use of narrow-band imaging for differential diagnosis of small colorectal polyps. Gastrointest Endosc 2009;69:278-83. 15. Sano Y, Ikematsu H, Fu KI, et al. Meshed capillary vessels by use of narrow-band imaging for differential diagnosis of neoplastic and nonneoplastic colonic polyps. Am J Gastroenterol 2006;101:2711-6. 16. Katagiri A, Fu KI, Sano Y, et al. Narrow band imaging with magnifying colonoscopy as diagnostic tool for predicting histology of early colorectal neoplasia. Aliment Pharmacol Ther 2008;27:1269-74. 17. Machida H, Sano Y, Hamamoto Y, et al. Narrow-band imaging in the diagnosis of colorectal mucosal lesions: a pilot study. Endoscopy 2004;36:1094-8. 18. Wada Y, Kudo SE, Kashida H, et al. Diagnosis of colorectal lesions with the magnifying narrow-band imaging system. Gastrointest Endosc 2009;70:522-31. 19. East JE, Suzuki N, Saunders BP. Comparison of magnified pit pattern interpretation with narrow band imaging versus chromoendoscopy for diminutive colonic polyps: a pilot study. Gastrointest Endosc 2007; 66:310-6. 20. My Su, Hsu CM, Ho YP, et al. Comparative study of conventional colonoscopy, chromoendoscopy, and narrow-band imaging systems in differential diagnosis of neoplastic and nonneoplastic colonic polyps. Am J Gastroenterol 2006;101:2711-6. 21. East JE, Suzuki N, Bassett P, et al. Narrow band imaging with magnification for the characterization of small and diminutive colonic polyps: pit pattern and vascular pattern intensity. Endoscopy 2008;40:811-7. 22. Chiu HM, Chang CY, Chen CC, et al. A prospective comparative study of narrow-band imaging, chromoendoscopy, and conventional colonoscopy in the diagnosis of colorectal neoplasia. Gut 2007;39:1092-6.
632 GASTROINTESTINAL ENDOSCOPY Volume 78, No. 4 : 2013
Hayashi et al 23. Tischendorf JJ, Wasmuth HE, Koch A, et al. Value of magnifying chromoendoscopy and narrow band imaging (NBI) in classifying colorectal polyps: a prospective controlled study. Endoscopy 2007;39:1092-6. 24. Chang CC, Hsieh CR, Lou HY, et al. Comparative study of conventional colonoscopy, magnifying chromoendoscopy, and magnifying narrow-band imaging system in the differential diagnosis of small colonic polyps between trainee and experienced endoscopist. Int J Colorectal Dis 2009;24:1413-9. 25. Singh R, Nordeen N, Mei SL, et al. West meets East: preliminary results of narrow band imaging with optical magnification in the diagnosis of colorectal lesions: a multicenter Australian study using the modified Sano’s classification. Dig Endosc 2011;(23 Suppl 1):126-30. 26. Rex DK. Narrow-band imaging without optical magnification for histologic analysis of colorectal polyps. Gastroenterology 2009;136: 1174-81. 27. Rex DK, Helbig CC. High yields of small and flat adenomas with highdefinition colonoscopes using either white light or narrow band imaging. Gastroenterology 2007;133:42-7. 28. Rastogi A, Bansal A, Wani S, et al. Narrow-band imaging colonoscopyd a pilot feasibility study for the detection of polyps and correlation of surface patterns with polyp histologic diagnosis. Gastrointest Endosc 2008;67:280-6. 29. Rastogi A, Pondugula K, Bansal A, et al. Recognition of surface mucosal and vascular patterns of colon polyps by using narrow-band imaging: interobserver and intraobserver agreement and prediction of polyp histology. Gastrointest Endosc 2009;69:716-22. 30. Rastogi A, Keighley J, Singh V, et al. High accuracy of narrow band imaging without magnification for the real-time characterization of polyp histology and its comparison with high-definition white light colonoscopy: a prospective study. Am J Gastroenterol 2009;104: 2422-30. 31. Rogart JN, Jain D, Siddiqui UD, et al. Narrow-band imaging without high magnification to differentiate polyps during real-time colonoscopy: improvement with experience. Gastrointest Endosc 2008;68: 1136-45. 32. Sikka S, Ringold DA, Jonnalagadda S, et al. Comparison of white light and narrow band high definition images in predicting colon polyp histology, using standard colonoscopes without optical magnification. Endoscopy 2008;40:818-22. 33. Ignjatovic A, East JE, Guenther T, et al. What is the most reliable imaging modality for small colonic polyp characterization? Study of white-light, autofluorescence, and narrow-band imaging. Endoscopy 2011;43:94-9. 34. Zhon QJ, Yang JM, Fei BY, et al. Narrow-band imaging endoscopy with and without magnification in diagnosis of colorectal neoplasia. World J Gastroenterol 2011;17:666-70. 35. Lee CK, Lee SH, Hwangbo Y. Narrow-band imaging versus I-Scan for the real-time histological prediction of diminutive colonic polyps: a prospective comparative study by using the simple unified endoscopic classification. Gastrointest Endosc 2011;74:603-9. 36. Rastogi A, Early DS, Gupta N, et al. Randomized, controlled trial of standard-definition white-light, high-definition white- light, and narrow-band imaging colonoscopy for the detection of colon polyps and prediction of polyp histology. Gastrointest Endosc 2011;74: 593-602. 37. Hewett DG, Kaltenbach T, Sano Y, et al. Validation of a simple classification system for endoscopic diagnosis of small colorectal polyps using narrow-band imaging. Gastroenterology 2012;143: 599-607. 38. Bosman FT, Carneiro F, Hruban RH, et al, editors. WHO classification of tumours of the digestive system. Lyon (France): IARC Press; 2010. p. 160-5. 39. Milholland AV, Wheeler SG, Heieck JJ. Medical assessment by a Delphi group opinion technic. N Engl J Med 1973;288:1272-5. 40. Rex DK, Ahnen DJ, Baron JA, et al. Serrated lesions of the colorectum: review and recommendations from an expert panel. Am J Gastroenterol 2012;107:1315-29.
www.giejournal.org