Diagnosis of colorectal lesions with a novel endocytoscopic classification a pilot study

Transcription

1 Original article 869 Diagnosis of colorectal lesions with a novel endocytoscopic classification a pilot study Authors S-E Kudo 1, K. Wakamura 1, N. Ikehara 1,Y.Mori 1, H. Inoue 1, S. Hamatani 2 Institutions 1 Digestive Disease Center, Showa University Northern Yokohama Hospital, Yokohama, Japan 2 Division of Pathology, Showa University Northern Yokohama Hospital, Yokohama, Japan submitted 18 October 2010 accepted after revision 15 May 2011 Bibliography DOI /s Published online: Endoscopy 2011; 43: Georg Thieme Verlag KG Stuttgart New York ISSN X Corresponding author S.-E. Kudo, MD, PhD Digestive Disease Center Showa University Showa University Northern Yokohama Hospital 35-1 Chigasaki-chuo, Tsuzuki Yokohama Japan Fax: kudos@med.showa-u.ac.jp Background and study aims: Recent advances in endocytoscopy have enabled in vivo evaluation not only of structural atypia, but also of cellular atypia with observation of lumens and nuclei in the surface layer of the mucosa. The aim of this prospective pilot study was to evaluate the usefulness of our novel endocytoscopic classification in colorectal lesions. Patients and methods: A total of 206 consecutive patients were enrolled in the study and underwent endocytoscopic examination. Endocytoscopic images were stored electronically and two endoscopists blinded to the findings at live examination assigned them diagnoses using the endocytoscopic (EC) classification. The endocytoscopic diagnosis was then compared to the final histopathological diagnosis. Results: In all, 196 patients with 213 specimens were available for analysis. All normal mucosae were classified as EC1a and all hyperplastic Introduction The development of the magnifying scope has greatly benefited the field of colorectal endoscopic diagnosis, making possible in vivo observation of the pit patterns of tumor surfaces, thereby improving the accuracy and objectivity of qualitative lesion diagnosis [1]. As a result of these developments, pit pattern diagnosis has become an indispensable tool for colonoscopic diagnostics [2]. With the recent invention of ultra-magnifying endocytoscopy, however, this situation is changing radically. The results obtained from endocytoscopy correspond well with those of microscopic findings. Endocytoscopy has made it possible to diagnose living tumor cells in vivo [3,4] and to obtain an ultra-magnification pathological image simply by applying the scope to the target mucosa during an endoscopic examination. polyps as EC1b. Dysplasias were mainly classified as EC2, while massively invasive submucosal cancers (SMm) or worse, which have the possibility of metastasis, were mainly EC3b. Assuming that an EC1b classification was diagnostic of hyperplastic polyps, we were able to differentiate nonneoplastic from neoplastic lesions with a sensitivity of 100 % and a specificity of 100 % (P < 0.05). Assuming that an EC3b classification was diagnostic of SMm or worse, we were able to differentiate SMm or worse from other neoplastic lesions (dysplasias and slightly invasive submucosal cancers) with a sensitivity of 90.1 % and a specificity of 99.2% (P < 0.05). Conclusions: The endocytoscopic classification was particularly useful for differentiating between neoplastic and nonneoplastic lesions and between SMm or worse and other neoplastic lesions, which in the case of colorectal neoplasms would help to determine treatment. There have been some reports on in vivo endoscopic cell observation relating to endocytoscopy. Hamou et al. first attempted to observe cells using a rigid scope (Karl Storz GmbH & Co. KG, Tuttlingen, Germany) in 1980 and reported this as contact microscopy [5]. There also have been previous reports of endocytoscopy including one in the field of otolaryngology on cellular observation with contact endoscopy using a Karl Storz rigid scope [6], and observations of esophageal squamous epithelium by Kumagai et al. [7]. The efficacy of endocytoscopy in the esophagus has been shown in the clinical studies by Kumagai et al. [7] and Inoue et al. [3,8,9]. In the colorectal field, endocytoscopy has also been shown to be effective in differentiating between neoplastic and nonneoplastic lesions, as well as in the diagnosis of submucosal infiltrative cancers [4]. Rotondano et al. validated the classification proposed by Sasajima et al. by investigating 52 colorectal lesions

2 870 Original article [10]. However, the number of target lesions seemed to be insufficient to validate that classification, and the classification seemed a little complicated and difficult for a trainee to learn. For this reason a new, simpler classification was developed; its validation is described here. The aim of the present study was to evaluate the usefulness of our endocytoscopic classification by comparison to the final histopathological findings. Fig. 1 Integrated-type endocytoscope. Patients and methods Patients This prospective pilot study took place at the Digestive Disease Center of Showa University Northern Yokohama Hospital from May 2005 to August A total of 206 consecutive Japanese patients were enrolled in our pilot study for the correlation of the endocytoscopic classification with the histopathological diagnosis. There were 143 male and 63 female patients. Mean patient age was 63.5 ± 11.7 years. Patients with familial adenomatous polyposis or with tumor associated with ulcerative colitis were excluded from the study. All patients underwent colonic examination with the endocytoscope and had their lesions treated endoscopically or surgically at the facility. Patients without any colorectal lesions were requested to undergo a biopsy from normal mucosa. Ethical approval was granted by the local ethics review committee and informed consent to participation in the clinical trial was obtained from all participating patients. Materials All examinations were performed with an integrated-type endocytoscope (XCF-260EC1; prototype from Olympus Co., Tokyo, Japan; " Fig. 1) which has adjustable scope hardness with an outside diameter of 13.6 mm at the distal end and a working length of 1330 mm. This scope makes it possible to perform conventional and ultra-magnifying endoscopic observation consecutively using one-touch switch operation without changing scope. This scope provides approximately 40 % of the pixels of high-definition colonoscopes currently on the market. Its image quality is the same as that of the CF-Q260AI (Olympus), but exact data are not officially revealed by Olympus. In endocytoscopic mode it has a magnification capability of 450; the depth of field is 50 μm and the field of view is μm. The principle of endocytoscopic mode is that of contact endoscopy and it is set up with a fixed focus. The scope tip is applied to the target areas of the lesion after staining with methylene blue and crystal violet. Ultra-magnified endocytoscopic images are then immediately obtained with the mode-changing switch on the handle. Features recognizable from endocytoscopic procedures include the morphology of the lumens of gland ducts in the epithelial superficial layer, the shape of epithelial cell nuclei stained with methylene blue, and the shape of epithelial cell cytoplasm stained with crystal violet. On the basis of the results of our previous pilot study (not published), endocytoscopic images were grouped into five classes [11] ( " Table 1). This novel endocytoscopic classification was developed from the procedure of reassessing endocytoscopic images which had been obtained from 75 lesions between 2003 and 2004 through a probe-type endocytoscope [4], with unblinded pathologists and endoscopists working together. In determining the classification, we also referred to histopathological tissue diagnoses: structural atypia and cellular atypia. Thus, endocytoscopic classification was defined as focusing on the morphology of lumens and the shape of nuclei in endocytoscopic images. Nonneoplastic lesions were classified as EC1a and EC1b, while neoplastic lesions as EC2, EC3a and EC3b. Details of the classification are as follows: Images showing roundish lumens and a regular pattern of uniformly sized fusiform nuclei were classified as EC1a ( " Fig. 2 a), which corresponds to normal colorectal mucosa. Images showing narrow, serrated lumens and a dense pattern of small roundish nuclei were classified as EC1b ( " Fig. 2 b), corresponding to hyperplastic polyps. Images showing neoplastic characteristics that correspond to histopathological tissue images ranging from low to high grade dysplasias were classified as EC2 ( " Fig. 2 c). This was recognized by slit - like smooth lumens and a regular pattern of fusiform or roundish nuclei. Irregular and rough lumens and a large number of roundish nuclei strongly stained with methylene blue were classified as EC3a ( " Fig. 2 d), corresponding to high grade dysplasias or slightly invasive submucosal cancers (SMs). Finally, images showing unclear gland formation and agglomeration of distorted nuclei strongly stained with methylene blue were classified as EC3b ( " Fig. 2 e). In this group, the lumens were often difficult to recognize. The majority Classification Endocytoscopic findings Histopathology EC1 a Roundish lumens Fusiform nuclei b Narrow serrated lumens Small roundish nuclei EC2 Slit-like smooth lumens Uniform fusiform or roundish nuclei EC3 a Irregular and rough lumens A large number of roundish nuclei b Unclear gland formation Agglomeration of distorted nuclei Normal mucosa Hyperplastic polyp Dysplasia High grade dysplasias or SMs cancer SMm cancer or worse Table 1 Endocytoscopic classification of lesions in the colorectum.

3 Original article 871 Fig. 2 a EC1a: roundish lumens and a regular pattern of uniformly sized fusiform nuclei. b EC1b: narrow serrated lumens and a dense pattern of small roundish nodules. c EC2: slit - like smooth lumens and a regular pattern of fusiform or roundish nuclei. d EC3a: irregular and rough lumens and a large number of roundish nuclei strongly stained with methylene blue. e EC3b: unclear gland formation and agglomeration of distorted nuclei strongly stained with methylene blue. of EC3b tumors that exhibited such characteristics were actually massively invasive submucosal cancers (SMm) or worse. Methods All the patients initially underwent colonoscopy in conventional endoscopic mode to detect lesions prior to examination with endocytoscopy. Any detected lesions were enrolled in the study regardless of size. Lesions were first washed carefully with water and images of conventional endoscopic views were obtained. Then, the lesions were stained with 1% methylene blue and 0.05 % crystal violet for approximately 10 seconds. Excessive stain was washed off to avoid overstaining the cells and observation in endocytoscopic mode was started to obtain images of the endocytoscopic views of the lesions. These images were obtained using the integrated image capture system of the endocytoscope and saved in the server that hosted the database in JPEG format with a pixel array of and 24-bit color. If no localized colorectal lesions were seen, observation with endocytoscopy and a biopsy of normal mucosa was taken as agreed in the signed informed consent procedure. Both conventional endoscopic and endocytoscopic observation were performed by either of two trained endoscopists (S.K., N.I.). Subsequently, endoscopic or surgical therapy was performed on the lesions and the resected specimens subjected to histopathological analysis. For this, specimens were stained with conventional hematoxylin and eosin, and histopathological diagnoses, based on the World Health Organization criteria, were made by a single experienced pathologist (S.H.) who was blind to the conventional endoscopic and endocytoscopic findings. In the next part of the study, conventional endoscopic images and endocytoscopic images of each lesion were downloaded separately from the server for analysis by two readers (K.W., Y.M.), trained endoscopists who were completely blind to the histopathological analysis. First, only conventional endoscopic images were randomly allocated to both the readers. They analyzed all the images from each target lesion and classified each lesion as nonneoplastic or dysplasia or SMs or SMm or worse. If the readers came to different diagnoses about the same lesion, the final diagnosis was agreed by discussion and consensus. Next, only endocytoscopic images of each lesion were randomly allocated to the same readers. All the images from each target lesion were analyzed and diagnosed according to the endocytoscopic classification through discussion. Four endoscopists (S.K., N.I., K.W., Y.M.) involved in the present study had experience in analyzing endocytoscopic images with the probe-type endocytoscopic model [4] from 2003 to The diagnoses from the conventional endoscopic images and endocytoscopic images were compared to the final histopathological diagnoses, and the sensitivity, specificity, and diagnostic accuracy of both conventional endoscopic and endocytoscopic findings were assessed by reference to the histopathology. The diag-

4 872 Original article sm1a Slightly invasive (SMs) sm1b Massively invasive (SMm) sm1c Fig. 3 Classification of the degree of submucosal invasion [13]. SMs, slightly invasive submucosal cancer; SMm, massively invasive submucosal cancer. B A sm2 sm1: upper 1/3 sm1a: B/A 1/4 sm1b: B/A 1/4 1/2 sm1c: B/A 1/2 sm2: middle 1/3 sm3: lower 1/3 sm3 Table 2 Correlation between endocytoscopic diagnosis and histopathological diagnosis in 213 colorectal specimens. Endocytoscopic diagnosis Normal mucosa Hyperplastic polyp Histopathological diagnosis Dysplasia Invasive cancer Low grade High grade SMs SMm or worse EC 1a 9 EC 1b 10 EC EC3a EC3b 1 64 nostic efficacy of conventional endoscopic and endocytoscopic findings was also evaluated in the differential diagnoses: between neoplastic and nonneoplastic lesions, and between SMm or worse (for which surgery is indicated) and other neoplastic lesions (dysplasias and SMs cancers, for which endoscopic treatment is indicated). In a substudy, inter- and intraobserver agreement in diagnosing colorectal lesions using the endocytoscopic classification was calculated for the two readers (K.W., Y.M.). For this purpose, 50 images of endocytoscopic views were randomly selected from among all the target lesions and arranged randomly for assessment by the two readers. Two weeks after the initial assessment, the same images were randomly allocated to the readers again for assessment. Interobserver agreement was calculated from the results of the first reading and intraobserver agreement was determined by comparing the first and the second assessments. We used the Kudo classification [12] for the degree of submucosal invasion ( " Fig. 3). Sm1a or 1b cancer without vessel permeation does not metastasize. In contrast, a substantial proportion of sm1c, sm2, and sm3 lesions (approximately 10 %) show nodal metastasis [13]. We defined the former lesions as SMs and the latter lesions as SMm. In cases of endoscopically resected specimens, sm1 cancer was defined as a lesion infiltrating the submucosal layer but limited to the upper half of the resected submucosal layer, and sm2 and sm3 cancer were defined as lesions infiltrating deeper than halfway through the resected submucosal layer, according to the methods reported by Tsuruta et al. [14] The SPSS for Windows Version 11.0 statistical software package (SPSS Inc., Chicago, Illinois, USA) was used for data analysis. Fisher s exact test was used for assessment of the diagnostic efficacy of conventional endoscopic and endocytoscopic findings and unweighted κ statistics were calculated for both inter- and intraobserver agreement on endocytoscopic images. In all cases, mean values were expressed as mean ± standard deviation and P values less than 0.05 were considered statistically significant. Results All in all, 5772 endocytoscopic images from 223 lesions were obtained (mean ± SD 25.9 ± 25.3 images for each lesion), of which 327 images were unclear and unsuitable for analysis. Clear endocytoscopic images could not be obtained in 10 patients with 10 lesions, which were excluded from the analysis. In four of these cases, the lesions were covered with too much mucus to wash away, and six lesions could not be observed in the depressed areas due to their peripheral elevations. Thus, endocytoscopic images clear enough for analysis were obtained in 95.5% of cases. There were nine patients without any localized colorectal lesions, and all of these agreed to a biopsy from normal mucosa. Thus, 196 patients with 213 specimens were available for analysis; these 213 specimens were made up of 9 normal mucosae, 10 hyperplastic polyps (all traditional hyperplastic polyps, no sessile serrated adenomas/polyps or serrated adenomas), 114 dysplasias, 9 SMs cancers, and 71 SMm or worse cancers (including 29 advanced cancers). The mean size of lesion was 17.6 ± 12.9 mm. Lesions were located in the rectum in 64 cases, in the left colon in 71 cases, and in the right colon in 69 cases. No complications were caused by observation with endocytoscopy. The mean duration of endocytoscopic observation was 17.1 ±10.0 min, while total procedure time including insertion and treatment was 54.1 ± 21.0 min.

5 Original article 873 In this prospective pilot study, the positive predictive values of histopathological tissue diagnosis based on the endocytoscopic classification were 100 % for EC1a (diagnosing normal mucosa), 100 % for EC1b (diagnosing hyperplastic polyps), 98.0 % for EC2 (diagnosing low to high grade dysplasias), 71.4% for EC3a (diagnosing high grade dysplasias or SMs cancers), and 98.5 % for EC3b (diagnosing SMm or worse ) ( " Table 2). There was excellent correlation between endocytoscopic classification and histopathological diagnosis. When we assumed that EC1b was diagnostic for hyperplastic polyps, we were able to differentiate between neoplastic and nonneoplastic lesions with a sensitivity of 100 % and a specificity of 100 % (P < 0.05, " Table 3). When we assumed that EC3b was diagnostic for SMm cancer, we were able to differentiate SMm or worse from other neoplastic lesions (dysplasias and SMs cancers) with a sensitivity of 90.1% and a specificity of 99.2 % (P < 0.05, " Table 4). The correlation between the prediction of the histopathological diagnosis based on conventional endoscopic findings and the final histopathological diagnosis is shown in " Table 5, which showed an overall accuracy of 82.8 %. Endocytoscopy showed significantly higher diagnostic accuracy for prediction of neoplastic and nonneoplastic lesions than conventional endoscopy (P = 0.015), while the sensitivity and the specificity were comparable to those of conventional endoscopy (P = and P = respectively; " Table 6). Concerning differentiation between SMm or worse and other neoplastic lesions, endocytoscopy showed significantly higher specificity and diagnostic accuracy (P < and P = respectively), while the sensitivity was comparable to that of conventional endoscopy (P = 0.606; " Table 7). In a substudy, interobserver agreement and intraobserver agreement for the endocytoscopic classification of colorectal lesions were assessed for the two readers. There was excellent interobserver agreement with a κ value of [95 % confidence interval (CI): ] and also excellent intraobserver agreement with a κ score of (95 % CI: ) for K.W. and (95 % CI: ) for Y.M. respectively. Discussion At present, pit pattern diagnosis with magnifying chromoendoscopy is used for tissue characterization and differential diagnosis in the colorectum. Kudo et al. [1,2,12] reported that magnifying colonoscopy provides an accurate and immediate assessment of the histology of colorectal lesions. When we detect a colorectal localized lesion in our facility, we do not usually take a biopsy specimen. Instead, magnifying colonoscopy with dye staining enables us to predict the histology of the lesion on site and thus enables us to avoid the cost, time, and risk of biopsy and repeated colonoscopy. Moreover, precise prediction of histology with magnifying colonoscopes can also avoid unnecessary polypectomy (as in the case of a hyperplastic polyp). In that sense, chromoendoscopy with magnification Table 3 Differential diagnosis between nonneoplastic and neoplastic colorectal lesions using endocytoscopy (n = 204). Endocytoscopic Histopathological diagnosis diagnosis Nonneoplastic Neoplastic EC1b 10 0 Others (EC2, EC3a, EC3b) P < Table 4 Differential diagnosis between SMm or worse and other neoplastic colorectal lesions using endocytoscopy (n = 194). Endocytoscopic diagnosis Histopathological diagnosis SMm or worse EC3b 64 1 Others (EC2, EC3a) Other neoplastic lesion (dysplasia or SMs) SMm, massively invasive submucosal cancer; SMs, slightly invasive submucosal cancer. P < has already replaced biopsy or unnecessary polypectomy, at least in Japan. Recently, some experience with endocytoscopy in both the esophagus and the colorectum has been reported [4,9,15,16]. Compared to magnifying chromoendoscopy, which is usually capable of 100 magnification, endocytoscopy has a magnification capability of 450, which can be used for in vivo endoscopic cell observation. Magnifying chromoendoscopic diagnosis is based on pit patterns, which indirectly reflect the morphology of crypts but never reflect cells or nuclei, whereas endocytoscopic observation captures not only structural atypia but also cellular atypia with observation of lumens and nuclei in the mucosal surface layer. For this reason endocytoscopy is closer to virtual biopsy than magnifying chromoendoscopy in this regard. There is another technique for in vivo endoscopic cell observation endomicroscopy. Endomicroscopy was developed to observe the esophageal and gastric mucosa, usually using a nondye technique, with a confocal laser microscopy unit, making real-time in vivo observation at the cellular level possible [17]. Kiesslich et al. [18] classified colorectal tissues into normal, regenerative, and neoplastic, and compared them with the underlying pathological diagnosis. However, this system requires intravenous injection of a fluorescent dye (fluorescein sodium) and has some limitations in the identification of nuclei. In this pilot study, we classified endocytoscopic images of colorectal lesions into five distinct groups. Our endocytoscopic observation was performed utilizing two patterns. First, luminal morphology was examined. Second, the shapes of epithelial cell nuclei were evaluated, as these reflect the cellular atypia in the Readers prediction Final histopathological diagnosis Nonneoplastic Dysplasia or SMs SMm or worse Nonneoplastic Dysplasia or SMs SMm or worse SMs, slightly invasive submucosal cancer. SMm, massively invasive submucosal cancer. Table 5 Correlation between prediction of histopathological diagnosis based on conventional endoscopic findings and final histopathological diagnosis in 204 colorectal lesions.

6 874 Original article Table 6 Comparison between conventional endoscopy and endocytoscopy in terms of differentiating nonneoplastic from neoplastic colorectal lesions. Sensitivity, % P value * Specificity, % P value * Accuracy, % P value * Conventional image 60.0 ( ) ( ) ( ) Endocytoscopic image 100 ( ) 100 ( ) 100 ( ) * P values were calculated for the difference between conventional endoscopy and endocytoscopy. 95%CI, 95% confidence interval. Table 7 Comparison between conventional endoscopy and endocytoscopy in terms of differentiating SMm or worse from other neoplastic lesions. Sensitivity, % P value * Specificity, % P value * Accuracy, % (95 %CI) Conventional image 85.9 ( ) ( ) < ( ) Endocytoscopic image 90.1 ( ) 99.2 ( ) 95.9 ( ) SMm, massively invasive submucosal cancer. * P values were calculated for the difference between conventional endoscopy and endocytoscopy. 95%CI, 95% confidence interval. P value * gland ducts of the epithelial superficial layer. This procedure was not necessarily difficult because the diagnostic method was comparatively simple. Sasajima et al. [4] have already reported the usefulness of endocytoscopy by considering (i) the pattern of the cellular arrangement; (ii) the size, shape, and arrangements of colonic glands (pits); (iii) the size and shape of the cells; (iv) the size and shape of the nuclei; and (v) the nuclear cytoplasmic ratio. However, this diagnostic procedure was a little complicated, so we developed it into a novel classification which is more easily understood even by a trainee. The adapted endocytoscopic classification requires no assignment of scores. It involves simply evaluating the lumens and nuclei of the target lesions, which might be intuitive in a certain sense, but its reproducibility was very high. Both intra- and intraobserver agreement for the endocytoscopic classification of colorectal lesions were excellent. We found that EC1a in this classification was characteristic of normal mucosa. EC1b was seen in hyperplastic polyps. EC2, EC3a, and EC3b were seen in neoplasia: EC2 was useful for diagnosis of dysplasia while EC3b was useful for diagnosing SMm or worse. " Table 2 suggests that there was excellent correlation between the endocytoscopic classification and the histopathological diagnosis. In addition, we found that endocytoscopy was particularly useful not only for differentiating between neoplastic and nonneoplastic lesions, but also for differentiating between SMm or worse and other neoplastic lesions ( " Table 3 and " Table 4). Furthermore, compared to conventional endoscopy, the endocytoscopic diagnosis showed higher specificity and diagnostic accuracy in respect of the differentiation between SMm or worse and other neoplastic lesions. SMm or worse lesions should be surgically treated from the beginning because they are associated with a significant possibility of metastasis, whereas other neoplastic lesions such as dysplasias and SMs lesions can be treated by endoscopy. Thus, the endocytoscopic classification can be very useful in determining the treatment of colorectal neoplasms. Some may argue that the characteristics of the EC3b category, which are unclear gland formation and agglomeration of distorted nuclei, indirectly reflect mucosal surface erosion or stromal reaction, which are often characteristics of SMm or worse. An integrated-type endocytoscope was used in the present study, and this may be the first report on it in the world. Sasajima et al. [4] and Rotondano et al. [10] have already reported on the usefulness of a probe - type endocytoscope in the colorectum which had a magnification capability of 450 and an observation area of μm. Compared to the probe - type endocytoscope, the primary advantage of the integrated type is probably ease of observation without having to exchange the probe. With the probe type, devices require two or more exchanges for staining or washing in addition to endocytoscopic observation, and also a distal cap attached to the tip of the colonoscope may be needed to fix the scope onto the target site. That is why the integrated type of endocytoscope may be easier and simpler to use than the probe type. In the present study, no complications were experienced. We used methylene blue and crystal violet for staining. Olliver et al. [19] reported DNA damage caused by methylene blue, but in the present study the amount of methylene blue was very small and exposure time was very short. Moreover, no report has been published which shows a clear relationship between DNA damage and carcinogenesis by exposure to methylene blue in vivo. The concentration of methylene blue was 1 % and the duration of exposure 10 s in the present study. Compared to the report on the optimal ex vivo staining conditions for endocytoscopy by Kodashima et al. [20], our exposure time was shorter, but in vivo lesions stained more easily than ex vivo. Kodashima et al. also reported that crystal violet was found to be unsuitable for use as a stain for endocytoscopic observation because nuclei were not observed under any dilution or exposure time with crystal violet. By contrast, in the present study staining was performed not only with methylene blue but also with crystal violet, since crystal violet provided clear contrast views of lumens and cytoplasm in the epithelial superficial layer. The main shortcomings in the present study were difficulty in obtaining a clear endocytoscopic view and the amount of time taken for endocytoscopic observation. Although lesions were washed very carefully before staining, a clear endocytoscopic view was possible in only 95.5% of the cases, which was mainly due to mucus and the macroscopic form of the lesions: depressed lesions with peripheral elevations were difficult to observe in contact. Investigation of a more effective washing method is also needed. In addition to this drawback, endocytoscopic observation took approximately 17 minutes which sometimes might cause peristalsis of the colon or make patients feel uncomfortable. The time taken for endocytoscopic observation needs to be shortened by some newly developed device or further training. This study also has some limitations. The number of endoscopies was small despite the long study period. This is because there was just one integrated-type endocytoscope in the facility and

7 Original article 875 only two endoscopists who performed the endocytoscopy in the study. Another limitation relates to the question of whether endocytoscopic diagnosis was performed in vivo or not. In the present study, the endocytoscopic images were obtained in vivo but not actually diagnosed in vivo. This is because we wanted to assess the accuracy of diagnosis solely on the basis of the endocytoscopic images, in a manner blinded to the conventional endoscopic images, although we routinely perform in vivo endocytoscopic diagnosis in the facility. In addition, we did not measure interobserver agreement between endoscopists and pathologists in the present study. This is because in our routine work endoscopists perform endocytoscopic diagnosis themselves, without input from pathologists. However, we had reassessed endocytoscopic images together with unblinded pathologists and endoscopists when we were developing the endocytoscopic classification. In conclusion, endocytoscopy was particularly useful for the differentiation between neoplastic and nonneoplastic lesions and between SMm or worse and other neoplastic lesions, which would help to determine the treatment of colorectal neoplasms. This study was a pilot study without a control group or randomization. Evaluation of this classification with a higher evidence level is currently being carried out in an ongoing randomized trial. Acknowledgments We would like to express our gratitude to Mr. Amyn Haji, MA MBBChir MSc FRCS, London, for his invaluable support in the correction of this manuscript. Competing interests: None References 1 Kudo S, Hirota S, Nakajima T et al. Colorectal tumours and pit pattern. J Clin Pathol 1994; 47: Kudo S, Rubio CA, Teixeira CR et al. Pit pattern in colorectal neoplasia: endoscopic magnifying view. Endoscopy 2001; 33: Inoue H, Kazawa T, Sato Y et al. In vivo observation of living cancer cells in the esophagus, stomach, and colon using catheter-type contact endoscope, Endo Cytoscopy system. Gastrointest Endosc Clin N Am 2004; 14: , x -xi 4 Sasajima K, Kudo SE, Inoue H et al. Real-time in vivo virtual histology of colorectal lesions when using the endocytoscopy system. Gastrointest Endosc 2006; 63: Hamou J, Salat-Baroux J, Coupez F, De Brux J. Microhysteroscopy: a new approach to the diagnosis of cervical intraepithelial neoplasia. Obstet Gynecol 1984; 63: Andrea M, Dias O, Santos A. Contact endoscopy of the vocal cord: normal and pathological patterns. Acta Otolaryngol 1995; 115: Kumagai Y, Monma K, Kawada K. Magnifying chromoendoscopy of the esophagus: in-vivo pathological diagnosis using an endocytoscopy system. Endoscopy 2004; 36: Inoue H, Kudo SE, Shiokawa A. Technology insight: Laser-scanning confocal microscopy and endocytoscopy for cellular observation of the gastrointestinal tract. Nat Clin Pract Gastroenterol Hepatol 2005; 2: Inoue H, Sasajima K, Kaga M et al. Endoscopic in vivo evaluation of tissue atypia in the esophagus using a newly designed integrated endocytoscope: a pilot trial. Endoscopy 2006; 38: Rotondano G, Bianco MA, Salerno R et al. Endocytoscopic classification of preneoplastic lesions in the colorectum. Int J Colorectal Dis 2010; 25: Kudo S-E, Wakamura K, Kashida H, Ikehara N. Real time diagnosis in the colorectum with integrated type endocytoscope. Gastrointest Endosc 2010; 71: AB Kudo S, Tamura S, Nakajima T et al. Diagnosis of colorectal tumorous lesions by magnifying endoscopy. Gastrointest Endosc 1996; 44: Kudo S, Tamegai Y, Yamano H et al. Endoscopic mucosal resection of the colon: the Japanese technique. Gastrointest Endosc Clin N Am 2001; 11: Tsuruta O, Toyonaga A, Ikeda H, Tanikawa K. Clinicopathological study of superficial-type invasive carcinoma of the colorectum: special reference to lymph node metastasis. Int J Oncol 1997; 10: Inoue H, Kudo SE, Shiokawa A. Novel endoscopic imaging techniques toward in vivo observation of living cancer cells in the gastrointestinal tract. Dig Dis 2004; 22: Kudo SE, Takemura O, Ohtsuka K. Flat and depressed types of early colorectal cancers: from East to West. Gastrointest Endosc Clin N Am 2008; 18: , xi 17 Sakashita M, Inoue H, Kashida H et al. Virtual histology of colorectal lesions using laser- scanning confocal microscopy. Endoscopy 2003; 35: Kiesslich R, Burg J, Vieth M et al. Confocal laser endoscopy for diagnosing intraepithelial neoplasias and colorectal cancer in vivo. Gastroenterology 2004; 127: Olliver JR, Wild CP, Sahay P et al. Chromoendoscopy with methylene blue and associated DNA damage in Barrett s oesophagus. Lancet 2003; 362: Kodashima S, Fujishiro M, Takubo K et al. Ex - vivo study of high-magnification chromoendoscopy in the gastrointestinal tract to determine the optimal staining conditions for endocytoscopy. Endoscopy 2006; 38: