ASGE Technology Committee systematic review and meta

C0h

5

ASGE TECHNOLOGY COMMITTEESYSTEMATIC REVIEW AND META-ANALYSIS

opyright ª 2015 by the016-5107/$36.00ttp://dx.doi.org/10.1016

02 GASTROINTESTIN

ASGE Technology Committee systematic review and meta-analysisassessing the ASGE PIVI thresholds for adoptingreal-time endoscopic assessment of the histology ofdiminutive colorectal polyps

Abstract: In vivo real-time assessment of the histology of diminutive (%5 mm) colorectal polyps detected at co-lonoscopy can be achieved by means of an “optical biopsy” by using currently available endoscopic technologies.This systematic review and meta-analysis was performed by the American Society for Gastrointestinal Endoscopy(ASGE) Technology Committee to specifically assess whether acceptable performance thresholds outlined by anASGE Preservation and Incorporation of Valuable endoscopic Innovations (PIVI) document for clinical adoption ofthese technologies have been met. We conducted direct meta-analyses calculating the pooled negative predictivevalue (NPV) for narrow-band imaging (NBI), i-SCAN, and Fujinon Intelligent Color Enhancement (FICE)–assistedoptical biopsy for predicting adenomatous polyp histology of small/diminutive colorectal polyps. We also calcu-lated the pooled percentage agreement with histopathology when assigning postpolypectomy surveillance inter-vals based on combining real-time optical biopsy of colorectal polyps 5 mm or smaller with histopathologicassessment of polyps larger than 5 mm. Random-effects meta-analysis models were used. Statistical heterogeneitywas evaluated by means of I2 statistics. Our meta-analyses indicate that optical biopsy with NBI, exceeds the NPVthreshold for adenomatous polyp histology, supporting a “diagnose-and-leave” strategy for diminutive predictednonneoplastic polyps in the rectosigmoid colon. The pooled NPV of NBI for adenomatous polyp histology by us-ing the random-effects model was 91% (95% confidence interval [CI], 88–94). This finding was associated with ahigh degree of heterogeneity (I2 Z 89%). Subgroup analysis indicated that the pooled NPV was greater than 90%for academic medical centers (91.8%; 95% CI, 89-94), for experts (93%; 95% CI, 91-96), and when the optical bi-opsy assessment was made with high confidence (93%; 95% CI, 90-96). Our meta-analyses also indicate that theagreement in assignment of postpolypectomy surveillance intervals based on optical biopsy with NBI of diminu-tive colorectal polyps is 90% or greater in academic settings (91%; 95% CI, 86-95), with experienced endoscopists(92%; 95% CI, 88-96) and when optical biopsy assessments are made with high confidence (91%; 95% CI, 88-95).Our systematic review and meta-analysis confirms that the thresholds established by the ASGE PIVI for real-timeendoscopic assessment of the histology of diminutive polyps have been met, at least with NBI optical biopsy, withendoscopists who are expert in using this advanced imaging technology and when assessments are made withhigh confidence. (Gastrointest Endosc 2015;81:502-16.)

The American Society for Gastrointestinal Endoscopy(ASGE) Technology Committee periodically performs sys-tematic reviews and meta-analyses to evaluate endo-scopic technologies to determine whether these have metpreviously established Preservation and Incorporationof Valuable endoscopic Innovations (PIVI) thresholds.

A subcommittee of the ASGE Technology Committee,comprising committee members chosen for their individ-ual expertise, invited outside expert(s) in the subjectarea, and the Technology Committee Chair performedthe systematic review and meta-analysis. The results are

American Society for Gastrointestinal Endoscopy

/j.gie.2014.12.022

AL ENDOSCOPY Volume 81, No. 3 : 2015

then reviewed and approved by the entire TechnologyCommittee. The systematic review and meta-analysisare ultimately submitted to the ASGE Governing Boardfor approval. The systematic review and meta-analysisundergo peer review by outside experts in statistics andmeta-analysis before receiving final ASGE GoverningBoard approval.

The PIVI initiative is an ASGE program, the objectivesof which are to identify important clinical questionsrelated to endoscopy and to establish a priori diagnosticand/or therapeutic thresholds for endoscopic technologiesdesigned to resolve these clinical questions. Once endo-scopic technologies meet an established PIVI threshold,those technologies are appropriate to incorporate intoclinical practice, presuming the appropriate training in

www.giejournal.org

http://dx.doi.org/10.1016/j.gie.2014.12.022

http://www.giejournal.org

Assessing PIVI thresholds for real-time assessment of colorectal polyp histology

that endoscopic technology has been achieved. ASGE en-courages and supports the appropriate use of technolo-gies that meet its established PIVI thresholds.

INTRODUCTION

The majority of colorectal polyps found at screening co-lonoscopy are diminutive (%5 mm). These polyps seldomharbor advanced histological features (villous features orhigh-grade dysplasia) and very rarely harbor cancer. How-ever, based on current management guidelines, endoscop-ists encountering diminutive polyps are obliged to removethem and submit them to histopathology to determine thenext surveillance colonoscopy interval based on the histol-ogy of these polyps (adenomatous vs hyperplastic).1,2 Thecosts associated with resection and pathology evaluationof these diminutive polyps add substantially to the totalcost of colonoscopy, especially in countries such as theUnited States, where colonoscopy is a common modalityfor colorectal cancer screening.3

If a sufficiently accurate in vivo assessment of the histol-ogy of these diminutive polyps can be achieved by meansof an “optical biopsy,” this may allow for a paradigm shift inthe assessment and management of these polyps, whichcould significantly reduce the total cost of colonoscopywithout affecting its efficacy in reducing future risk of colo-rectal cancer. This paradigm shift would incorporate theadoption of a “diagnose-and-leave” strategy, in which theendoscopist leaves in situ diminutive rectosigmoid hyper-plastic polyps, and a “resect-and-discard” strategy, in whichdiminutive adenomatous polyps are resected after endo-scopic assessment of histology to allow determination ofsurveillance colonoscopy intervals, but not submitted forhistopathology evaluation.

Several in vivo endoscopic technologies exist that allowfor real-time characterization of diminutive colorectalpolyps that are superior to that achievable with white-light endoscopy. Electronic chromoendoscopy technolo-gies facilitate polyp characterization by enhancing thesurface pit pattern and highlighting the microvasculatureof these polyps. These technologies include narrow-bandimaging (NBI) (Olympus, Tokyo, Japan), i-SCAN (Pentax,Tokyo, Japan), and Fujinon Intelligent Color Enhancement(FICE) (Fujinon Inc, Saitama, Japan).4 Confocal laser endo-microscopy (CLE) has also been evaluated for this purpose.

The American Society for Gastrointestinal Endoscopy(ASGE) created a new initiative in 2011 entitled Preserva-tion and Incorporation of Valuable endoscopic Innovations(PIVI). The key objectives of the PIVI initiative are to iden-tify important clinical questions related to endoscopy andto establish a priori, diagnostic, and/or therapeutic thresh-olds for endoscopic technologies designed to resolve theseclinical questions. The ASGE has identified endoscopicpolyp characterization as a key area for new endoscopictechnologies and has outlined in a PIVI document entitled

www.giejournal.org

“The ASGE PIVI on Real-Time Endoscopic Assessment ofthe Histology of Diminutive Colorectal Polyps” the perfor-mance thresholds that these technologies should meetbefore adoption into clinical practice.5 Two thresholdshave been established in this PIVI document:1. For a technology to be used to guide the decision to

leave suspected rectosigmoid hyperplastic polyps5 mm or smaller in place (without resection), the tech-nology should provide a 90% or greater negative predic-tive value (NPV) (when used with high confidence) foradenomatous histology.

2. For colorectal polyps 5 mm or smaller to be resectedand discarded without pathologic assessment, endo-scopic technology (when used with high confidence)used to determine histology of these polyps, when com-bined with the histopathologic assessment of polypslarger than 5 mm, should provide 90% or greater agree-ment in assignment of postpolypectomy surveillance in-tervals when compared with decisions based onpathology assessment of all identified polyps.The systematic review and meta-analyses were per-

formed by the ASGE Technology Committee to specificallyassess whether these PIVI thresholds have been met, basedon the existing literature. Input was also sought from thechair (D.K.R.) of the ASGE Committee that wrote the orig-inal PIVI document.

METHODS

Data sources and search strategiesA comprehensive search of several English-language da-

tabases was conducted for studies published betweenJanuary 1, 1980 and May 14, 2014. The databases includedOvid MEDLINE In-Process & Other Non-Indexed Citations,Ovid MEDLINE, Ovid Embase, Ovid Cochrane Central Reg-ister of Controlled Trials, Ovid Cochrane Database of Sys-tematic Reviews, and Web of Science. The searchstrategy was designed and conducted by an experiencedlibrarian with input from the study team. Controlled vocab-ulary supplemented with keywords was used to searchfor studies evaluating electronic chromoendoscopy (NBI,i-SCAN, and FICE) for the characterization of colonicpolyps. The detailed search strategy is available uponrequest. Relevant studies were also identified from thebibliography of studies obtained through the search. Asimilar process was also used for studies evaluatingconfocal laser endomicroscopy (CLE) for the characteriza-tion of colonic polyps.

Study selectionWe reviewed titles and abstracts of studies retrieved by

our search strategy for potential eligibility for inclusion inthe meta-analysis. Based on the initial review of study titlesand abstracts, we narrowed the search to 49 relevantfull-length studies evaluating NBI and 25 studies evaluating

Volume 81, No. 3 : 2015 GASTROINTESTINAL ENDOSCOPY 503



i-SCAN and FICE. We reviewed the full text of thesestudies, and included in the meta-analysis:1. Studies that included data on the NPV of real-time NBI,

i-SCAN, or FICE for adenomatous histology of small anddiminutive colorectal polyps detected during colonos-copy and/or

2. Studies that assessed agreement in assigning postpoly-pectomy surveillance intervals when combining real-time NBI, i-SCAN, or FICE in situ optical biopsy ofcolorectal polyps 5 mm or smaller along with histopath-ologic assessment of polyps larger than 5 mm comparedwith decisions based on histopathologic assessmentalone of all identified polyps.Two reviewers performed study selection (B.K.A.D. and

N.T.); when a disagreement occurred, a third blindedreviewer (S.B.) was consulted to resolve the disagreement.Both reviewers eventually agreed on all included studies.For inclusion in the meta-analysis, a study had to meet thefollowing inclusion criteria: human trial, published in En-glish (full-text) in a peer-reviewed journal, and evaluatingthe ability of NBI, i-SCAN, or FICE in performing real-timein situ optical biopsy of small and diminutive colorectalpolyps, compared with the criterion standard of histopathol-ogy, in achieving 1 or both of the PIVI thresholds. Abstracts,letters, editorials, expert opinions, reviews without originaldata, case reports, and studies not directly assessing at least1 of the 2 PIVI thresholds were excluded. All prospective,randomized trials included in the meta-analysis met the ma-jority of the criteria set forth by the Evidence-Based Gastro-enterology Steering Group for methodologic quality,indicating studies of reasonable quality.6

A similar review process for CLE revealed only 5 pub-lished studies with a mix of endoscopy-based and probe-based CLE, reporting real-time classification of polyps.Given the variability of these studies in assessing the PIVIthresholds, we could not combine them in a meta-analysis.

Data extractionTwo independent reviewers performed data extraction

(B.K.A.D. and N.T.) from each selected citation. When am-biguity on outcomes determination was present, a thirdreviewer (S.B.) was consulted, and the outcome was deter-mined by consensus. Data extracted included the year thestudy was published, the country where the study was con-ducted, setting (academic center vs community practice),expertise of the endoscopist in performing optical biopsyof colorectal polyps by using NBI, i-SCAN, or FICE (expertvs novice), training in optical NBI biopsy as part of thestudy protocol (yes vs no), criteria used for optical biopsy(vascular pattern intensity [VPI], Kudo, Sano, ModifiedNarrow Band Imaging International Colorectal EndoscopicClassification [M-NICE], NICE, or combination), use ofhigh-definition white-light endoscopy (yes vs no), use ofmagnification (yes vs no), processor type, high confidenceinterpretation of the optical biopsy (yes vs no), number ofsmall and diminutive polyps detected, number of polyps

504 GASTROINTESTINAL ENDOSCOPY Volume 81, No. 3 : 2015

per patient, percentage of polyps with adenomatous histol-ogy, NPV of NBI, i-SCAN, or FICE optical biopsy for predict-ing adenomatous polyp histology of small and diminutivecolorectal polyps, and percentage of agreement with histo-pathology when assigning postpolypectomy surveillanceintervals based on combining real-time NBI, i-SCAN, orFICE optical biopsy of colorectal polyps 5 mm or smalleralong with histopathologic assessment of polyps largerthan 5 mm in size.

Statistical analysisTo best summarize the available evidence, we conducted

direct meta-analyses calculating the pooled NPV with 95%confidence intervals (CI) for NBI-, i-SCAN-, and FICE-assisted optical biopsy for predicting adenomatous polyphistology of small and diminutive colorectal polyps. Wealso calculated the pooled percentage of agreement withhistopathology and 95% CI when assigning postpolypectomysurveillance intervals based on combining real-time NBI-,i-SCAN, and FICE-assisted optical biopsy of colorectal polyps5 mm or smaller with histopathologic assessment of polypslarger than 5 mm. Random-effects meta-analysis modelswere used. Statistical heterogeneity was evaluated by meansof the I2 statistic; an I2 value greater than 50% was consideredto indicate high statistical heterogeneity. A funnel plot and Eg-ger regression asymmetry were used to assess potential pub-lication bias. We also performed subgroup analysis to analyzethe effects of expertise, setting, and high confidence interpre-tation on the pooled primary outcomes. Analyses were per-formed by using Comprehensive Meta-analysis softwareversion 2 (Biostat, Inc, Englewood, NJ).

RESULTS

The search results are summarized in Figure 1. The liter-ature search captured a total of 771 citations for NBI and 159citations for i-SCAN and FICE. Review for citation duplicationeliminated 377 citations for NBI and 85 citations for i-SCANand FICE. Of the 74 remaining citations for i-SCAN and FICE,39 citations were for FICE and 35 citations were for i-SCAN.Title reviews led to the exclusion of 244 citations for NBI, 13citations for FICE, and 11 citations for i-SCAN. Of the subse-quently reviewed abstracts (150 for NBI, 26 for FICE, and 24for i-SCAN), 101, 14, and 11 citations, respectively, wereexcluded because they were not applicable to the currentmeta-analysis. The full-length manuscripts of the remaining49 NBI citations, 12 FICE citations, and 13 i-SCAN citationswere reviewed. Of these, 29 citations for NBI, 4 citationsfor FICE, and 5 citations for i-SCAN were excluded becausethey did not meet inclusion/exclusion criteria.

Twenty remaining citations were ultimately included inthe NBI meta-analysis, of which 19 evaluated the NPV ofoptical NBI biopsy and 10 evaluated agreement in postpoly-pectomy surveillance intervals. The 20 included studies werepublished between 2008 and 2014, with the majority (17/20)

www.giejournal.org


Ovid Medline : 204Ovid Embase: 306Ovid CCRCT: 43

Ovid CDSR: 5Web of Science: 213

Duplicate citations: 377

Duplicate citations: 85

Excluded (not relevant): 244

Excluded (not relevant): 24

Reason for exclusion: Not in English, did notevaluate adults ≥ 18yo, not original data, notdiminutive or small polyps , no real-timepolyps assessment, no-histopath ref, did notlook at colon polyps , IBD cohort, high riskhereditary syndromes, did not apply to keyPIVI questions, other technology used (CLE)

Reason for exclusion: Not in English, did notevaluate adults ≥ 18yo, not original data, notdiminutive or small polyps , no real-timepolyps assessment, no-histopath ref, did notlook at colon polyps , IBD cohort, high riskhereditary syndromes , did not apply to keyPIVI questions, other technology used (CLE)

Reasons for exlusion: Did not exclusivelyassess diminutive or small polyps, noreal-time polyps assessment, no-histopath ref, did not apply to key PIVIquestions

Reasons for exlusion: Did exclusivelyassess diminutive or small polyps , noreal-time polyps assessment, no-histopath ref, did not apply to key PIVIquestions

Ovid Medline : 44Ovid Embase: 64

Ovid CCRCT: 13Ovid CDSR: 1Web of Science: 37

771 Citations

159 Citations

394 CitationsRetrieved

Title Review

150 AbstractReview

49 Full ArticleReview

Assessing agreementin survillence intervals

10 citations3082 patients

Assessing NPV19 citations

4013 small ordiminutive polyps

74 CitationsRetrieved

Title Review

50 AbstractReview

25 Full ArticleReview

I-Scan8 citations

979 small or diminutivepolyps

FICE8 citations

1243 small ordiminutive polyps

A

B

Figure 1. Flow diagram depicting included studies selected for NBI meta-analysis (A) and i-SCAN and FICE meta-analyses (B). CLE, confocal laser endos-copy; FICE, Fujinon Intelligent Color Enhancement; IBD, inflammatory bowel disease; NPV, negative predictive value; PIVI, Preservation and Incorpora-tion of Valuable endoscopic Innovations.


of the studies conducted at academic medical centers.Twelve studies were performed by experts in optical NBI bi-opsy of colorectal polyps. Seventeen of the included studies

www.giejournal.org

used high-definition white-light endoscopy in addition toNBI. Eleven of the 20 included studies designating the con-fidence level of interpreting the optical biopsies (Table 1).



TABLE 1. Narrow-band imaging studies

Study (year), country

Setting(academic

vs community)Expertisein NBI Training

Criteria foroptical biopsy

Highdefinition Magnification

Highconfidence NPV

Surveillanceintervals

East et al (2008), UK Academic Yes No VPI Yes Yes No Yes No

Rogart et al (2008), USA Academic No Yes Kudo/VPI Yes Yes No Yes No

Ignjatovic et al (2009), UK Academic Mixed Yes VPI Yes No Yes Yes Yes

Rex et al (2009), USA Academic Yes Yes Kudo/VPI Yes On demand Yes Yes Yes

Sano et al (2009), Japan Academic Yes No Sano No Yes No Yes No

Van Den Broek et al (2009),the Netherlands

Academic No No Kudo/VPI No Yes No Yes No

Henry et al (2010), USA Academic Yes Yes Sano/Emura Yes On demand No Yes No

Lee et al (2011),South Korea

Academic Yes No Pit vascularpattern

Yes No Yes Yes No

Gupta et al (2012),USA

Academic Yes No Pit vascularpattern

Yes No No Yes Yes

Paggi et al (2012),Italy

Community Yes Yes Pit vascularpattern

Yes No Yes Yes Yes

Hewett et al (GIE)(2012), USA

Academic Yes No M-NICE Yes Yes Yes Yes No

Hewett et al(Gastroenterology)(2012), USA

Academic No Yes NICE Yes No Yes Yes No

Shahid et al (2012), USA Academic Yes No Sano Yes No No Yes No

Kuiper et al (2012), theNetherlands

Community No Yes Kudo Yes No Yes Yes Yes

Coe et al (2012), USA Academic No Yes Pit vascularpattern

Yes No No No Yes

Sakamoto et al (2012),Japan

Academic No Yes Sano No Yes No Yes No

Repici et al (2013), Italyand the Netherlands

Academic Yes Yes M-NICE Yes Yes Yes Yes Yes

Singh et al (2013),Australia and Japan

Academic Yes No Sano Yes Yes/dual focus Yes Yes Yes

Ladabaum et al (2013),USA

Community No Yes NICE Yes No Yes Yes Yes

Wallace et al (2014), USA Academic Yes Yes NICE Yes Yes Yes Yes Yes

Academic Yes Yes NICE Yes Yes/dual focus Yes Yes Yes

NBI, Narrow-band imaging; NPV, negative predictive value; VPI, vascular pattern intensity; NICE, Narrow Band Imaging International Colorectal Endoscopic Classification; M-NICE,Modified Narrow Band Imaging International Colorectal Endoscopic Classification.


Sixteen remaining citations were ultimately included inthe i-SCAN and FICE meta-analyses, of which 8 evaluatedi-SCAN and 8 evaluated FICE. The NPV of optical i-SCANor FICE biopsy was evaluated in all studies; however,only 2 studies evaluated agreement in postpolypectomysurveillance intervals for FICE and 1 for i-SCAN. The 16included studies were published between 2008 and 2014,with the majority (13/16) conducted at academic medicalcenters. Ten studies were performed by experts in opticalbiopsy of colorectal polyps. All included i-SCAN andFICE studies used high-definition white-light endoscopyin addition to the respective technology. Only 1 of the16 included i-SCAN/FICE studies designated the confi-dence level of interpreting the optical biopsies (Table 2).


Meta-analysis of NBI studies evaluating NPV foradenomatous polyp histology

Nineteen studies reported or provided informationenabling the calculation of the NPV of optical biopsy per-formed by using NBI for predicting adenomatous polyphistology of small and diminutive colorectal polyps.7-25

These studies collectively evaluated 4013 in situ small ordiminutive colorectal polyps in real-time by using NBIand compared it with criterion standard histopathology.The median prevalence of adenomas among these polypswas 48.5% (range 18%–88%). There was no evidence ofpublication bias based on a review of the funnel plot(Appendix 1, available online at www.giejournal.org). Thepooled NPV by using the random effects model was 91%

www.giejournal.org



TABLE 2. i-SCAN and FICE studies

Study (year), country

Setting(academic

vs community)

Expertisein i-SCAN or

FICE Training

Criteria foropticalbiopsy

Highdefinition Magnification

Highconfidence NPV

Surveillanceintervals

i-SCAN

Lee et al (2011), Korea Academic Yes No Kudo/VPI Yes No Yes Yes No

Hoffman et al 2 (2010),Germany

Academic Yes No Kudo/VPI Yes No No Yes No

Hoffman et al 1 (2010),Germany

Academic Yes No Kudo Yes No No Yes No

Chan et al (2012), USA Academic No Yes Kudo Yes No No Yes No

Hong 2 et al (2012)South Korea

Academic No No Kudo/VPI Yes No No Yes No

Hong 1 et al (2012),South Korea

Academic No No Kudo/VPI Yes No No Yes No

Pigo et al (2013), Italy Community Yes No NICE Yes No No Yes No

Schachschal et al (2014),Germany

Mixed No Yes Kudo Yes No No Yes Yes

FICE

Pohl et al (2008), Germany Community No Yes Kudo/VPI Yes No No Yes No

Togashi et al (2009),Japan

Academic No Yes Vascularpattern

Yes No No Yes No

Buchner et al (2010), USA Academic Yes Yes Kudo/VPI Yes No No Yes No

Dos Santo et al (2010),Brazil

Academic Yes No Vascularpattern

Yes Yes No Yes No

Kim et al (2011)South Korea

Academic Yes Yes Kudo/VPI Yes Yes No Yes No

Longcroft et al (2011),England


Yes No No Yes Yes

Dos Santo et al (2012),Brazil


Yes Yes No Yes No

Longcroft et al (2012),England


Yes Yes No Yes Yes

FICE, Fujinon Intelligent Color Enhancement; NPV, negative predictive value; VPI, vascular pattern intensity.


(95% CI, 88–94) (Fig. 2). This finding was associated withhigh degree of heterogeneity (I2 Z 89%). Therefore, weperformed multiple subgroup analyses to account forsome of the factors contributing to this high degree ofheterogeneity.1. Effect of practice setting: No significant difference was

noted in the pooled NPV of studies conducted at aca-demic medical centers (91.8%; 95% CI, 89-94) comparedwith community practices (88.3%, 95% CI, 82-94) (Fig. 3).

2. Effect of operator experience: A subgroup analysisbased on experience in interpreting real-time optical bi-opsies of colorectal polyps indicated that only expertsmet the PIVI threshold of a 90% or higher NPV with apooled NPV for experts of 93% (95% CI, 91-96). Thisfinding was associated with a lesser degree of heteroge-neity (I2 Z 78%) (Fig. 4). The NPV for novice operatorswas 87% (95% CI, 83-91).

3. Effect of confidence level: The pooled NPV was higher(93%; 95% CI, 90-96) when the optical biopsy

www.giejournal.org

assessment was made with high confidence comparedwith when no information on the confidence level wasprovided (88%; 95% CI, 84-92) (Fig. 5).

4. Effect of combined operator experience andconfidence level: Novice operators approached thePIVI NPV threshold when their optical biopsies wereperformed with high confidence (90%; 95% CI, 86-94).Experienced operators exceeded PIVI thresholds whenreporting assessments with high confidence (95%;95% CI, 92-98).

Meta-analysis of the degree of agreement inassignment of postpolypectomy surveillanceintervals based on NBI optical biopsy versusthose based on histopathology

Ten studies including 3082 patients reported on thedegree of agreement with histopathology when assigningpostpolypectomy surveillance intervals based oncombining real-time NBI optical biopsy of colorectal polyps



NPV for NBI Optical Biopsy

Study name

Ignjatovic 2009Rogart 2008

East 2008

Rex 2009Sano 2009

Van Den Broek 2009Henry 2010Lee 2011Gupta 2012

Hewett_1_2012Hewett_2_2012

Kuiper_2012Paggi_2012Sakamoto_2012Shahid_2012

Ladabaum_2013Repici_2013Singh_2013Wallace_1_2014Wallace_2_2014

Random

MeanLowerlimit

Upperlimit Total

94.081.082.3

95.490.090.290.7

92.095.4

99.495.086.686.462.275.091.4

92.0100.0

96.097.091.1

89.173.073.892.782.9

85.184.5

86.793.198.891.079.880.946.9

66.186.3

88.079.993.095.088.7

98.9

89.090.798.197.1

95.496.997.397.7

100.099.093.392.077.583.996.5

96.0100.099.099.093.6

96

265213

314

150206

90125516

20117823139927010321920440

10489

NPV 90%

Figure 2. Forest plot of studies evaluating the negative predictive value(NPV) for narrow-band imaging (NBI)–assisted optical biopsy for predict-ing adenomatous polyp histology of small/diminutive colorectal polyps.


5 mm or smaller with histopathologic assessment ofpolyps larger than 5 mm using the U.S. Multi-SocietyTask Force (MSTF) postpolypectomy surveillance inter-vals.7-12,15,18,20,26 The pooled percentage of agreementusing the random-effects model was 89% (95% CI, 85-93)(Fig. 6). This finding was associated with a significant de-gree of heterogeneity (I2 Z 93%). Therefore, we per-formed multiple subgroup analyses to account for someof the factors contributing to this high degree ofheterogeneity.1. Effect of practice setting: The pooled percentage of

agreement was higher for studies conducted in aca-demic medical centers (91% [95% CI, 86-95]) comparedwith community practices (82% [95% CI, 74-90])(Fig. 7).

2. Effect of operator experience: A subgroup analysisbased on endoscopist experience in interpreting real-time optical biopsies of colorectal polyps indicatedthat experts reached the PIVI threshold of 90% or higheragreement (92%; 95% CI, 88-96). This finding was asso-ciated with lesser degree of heterogeneity (I2 Z 72%)(Fig. 8). Experts outperformed novice endoscopists,who had a lower pooled percentage of agreement(82%; 95% CI, 75-88).

3. Effect of confidence level: The pooled percentage ofagreement was higher when the optical biopsy wasmade with high confidence (91%; 95% CI, 88-95)compared with when no information on the confidencelevel was provided (79%; 95% CI, 71-86) (Fig. 9).

4. Effect of combined operator experience and confidencelevel: Novice operators improved their agreement with


postpolypectomy surveillance intervals when their opticalbiopsies were performed with high confidence (87%;95% CI, 82-93). Experienced operators when reportingwith high confidence exceeded PIVI thresholds (93%;95% CI, 90-96).

Meta-analyses of i-SCAN and FICE studies inmeeting PIVI thresholds

i-SCAN. Eight studies reported or provided informa-tion enabling the calculation of the NPV of optical biopsiesperformed by using i-SCAN for predicting adenomatouspolyp histology of small and diminutive colorectalpolyps.14,27-32 These studies collectively evaluated 979small or diminutive colorectal polyps in real-time usingi-SCAN and compared their evaluation with the criterionstandard histopathology. There was no evidence of publi-cation bias based on review of the funnel plot (Appendix 2).The pooled NPV by using the random-effects model was84% (95% CI, 76–91). This finding was associated with asignificant degree of heterogeneity (I2 Z 95). Therefore,we performed a subgroup analyses based on endoscopistexperience in performing and interpreting optical biopsiesof colorectal polyps. Experienced endoscopists meet thePIVI NPV threshold for adenomatous histology of smalland diminutive colorectal polyps with a NPV of 96% (95%CI, 94-98) compared with a NPV of 72% (95% CI, 69-76)for novice endoscopists (Fig. 10). Only 1 i-SCAN study27

evaluated the degree of agreement with histopathologywhen assigning postpolypectomy surveillance intervalsbased on combining real-time i-SCAN optical biopsy ofcolorectal polyps 5 mm or smaller with histopathologicassessment of polyps larger than 5 mm in size using theMSTF postpolypectomy surveillance intervals. The re-ported degree of agreement of i-SCAN optical biopsywith histopathology in the assignment of postpolypectomysurveillance intervals based on this study was 69.5% (95%CI, 63-75), which did not meet the PIVI threshold.

Fujinon Intelligent Color Enhancement. Eightstudies reported or provided information enabling thecalculation of the NPV of optical biopsy performed byFICE for adenomatous polyp histology of small and dimin-utive colorectal polyps.33-40 These studies collectivelyevaluated 1243 small or diminutive colorectal polyps inreal-time by using FICE and compared the evaluationwith the criterion standard histopathology. There was noevidence of publication bias based on a review of the fun-nel plot (Appendix 3). The pooled NPV by using therandom-effects model was 80% (95% CI, 76–85). Thisfinding was associated with a significant degree of hetero-geneity (I2 Z 70). Subgroup analyses based on operatorexperience and use of magnification with the FICE opticalbiopsy indicated that operator experience did not improvethe NPV of FICE for predicting adenomatous polyp his-tology; however, the use of magnification did improveFICE performance with a NPV of 85% (95% CI, 79–91)(Fig. 11). Only 2 FICE studies33,35 evaluated the degree

www.giejournal.org


Subgroup by Practice Setting

Group byAcademic

Study name Statistics for each study

NoNoNoNoYesYesYesYesYesYesYesYesYesYesYesYesYesYesYesYesYesYes

Random

Random

Kuiper_2012Paggi_2012Ladabaum_2013

East 2008Rogart 2008Ignjatovic 2009Rex 2009Sano 2009Van Den Broek 2009Henry 2010Lee 2011Gupta 2012Hewett_1_2012Hewett_2_2012Sakamoto_2012Shahid_2012Repici_2013Singh_2013Wallace_1_2014Wallace_2_2014

Mean

86.686.491.488.394.081.082.395.490.090.290.792.095.499.495.062.275.092.0

96.097.091.8

100.0

Lowerlimit

Upperlimit

79.880.986.382.389.173.073.892.782.985.184.586.793.198.891.046.966.188.079.993.095.089.3

93.392.096.594.298.989.090.798.197.195.496.997.397.7

100.099.077.583.996.0

100.099.099.094.3

NPV 90%

Figure 3. Forest plot of studies evaluating the negative predictive value (NPV) for narrow-band imaging–assisted optical biopsy for predicting adenoma-tous polyp histology of small/diminutive colorectal polyps stratified by practice setting (academic vs community).

Kuiper_2012

Paggi_2012

Ladabaum_2013

East 2008

Rogart 2008Ignjatovic 2009

Rex 2009Sano 2009

Van Den Broek 2009

Henry 2010Lee 2011Gupta 2012Hewett_1_2012

Hewett_2_2012

Sakamoto_2012

Shahid_2012Repici_2013Singh_2013Wallace_1_2014Wallace_2_2014

NoNoNoNoNoNoNoNo

YesYesYes

YesYesYesYesYes

YesYes

YesYesYesYes

Group byExpert


MeanLowerlimit

Upperlimit

NPV 90%

Random

Random

81.082.390.295.086.662.291.487.394.095.490.090.792.095.499.486.475.092.0

100.096.097.093.2

73.073.885.191.079.846.986.383.389.192.782.984.586.793.198.880.966.188.079.993.095.090.6 95.8

99.099.0100.096.083.992.0

100.097.797.396.997.198.198.991.396.577.593.399.095.490.789.0

Subgroup by Endoscopist Expertise

Figure 4. Forest plot of studies evaluating the negative predictive value (NPV) for narrow-band imaging–assisted optical biopsy for predicting adenoma-tous polyp histology of small/diminutive colorectal polyps stratified by endoscopist expertise (expert vs novice).

www.giejournal.org Volume 81, No. 3 : 2015 GASTROINTESTINAL ENDOSCOPY 509



Statistics for each study

Subgroup by Confidence Level

Group byHigh Confidence

Study name

Kuiper_2012Paggi_2012Ladabaum_2013

East 2008Rogart 2008

Ignjatovic 2009Rex 2009

Sano 2009Van Den Broek 2009Henry 2010

Lee 2011

Gupta 2012

Hewett_1_2012Hewett_2_2012

Sakamoto_2012Shahid_2012

Repici_2013Singh_2013Wallace_1_2014Wallace_2_2014

NoNoNoNoNoNoNoNoNoYesYesYesYesYesYesYesYesYesYesYesYesYes

Random

Random

MeanLowerlimit

Upperlimit

NPV 90%

94.081.090.090.290.795.462.275.088.082.395.492.099.495.086.686.491.492.0

100.096.097.093.1

89.173.082.985.184.593.146.966.184.173.892.786.798.891.079.880.986.388.079.993.095.090.2

98.989.097.195.496.997.777.583.991.890.798.197.3

100.099.093.392.096.596.0

100.099.099.096.0

Figure 5. Forest plot of studies evaluating the negative predictive value (NPV) for narrow-band imaging (NBI)–assisted optical biopsy for predictingadenomatous polyp histology of small/diminutive colorectal polyps, stratified by high confidence interpretation of the optical biopsy.

Agreement with Surveillance Intervals Outcome forNBI Optical Biopsy

Study name

Kuiper 2012

Coe 2012

Paggi 2012

Ladabaum 2013

Ignjatovic 2009

Rex 2009

Gupta 2012

Repici 2013

Singh 2013

Wallace-1 2014

Wallace-2 2014

Random

MeanLowerlimit

Upperlimit Total 90% Agreement

95.00

94.12

86.10

85.2881.48

70.00

92.00

96.5579.90

95.00

94.0088.63

91.06

91.27

82.91

80.6972.44

65.32

88.02

94.1372.51

93.01

91.0184.57

98.94

96.97

89.29

89.8790.52

74.68

95.98

98.9787.29

96.99

96.9992.70

82

136

410

19754

317

212

871065

264

258

Figure 6. Forest plot of narrow-band imaging (NBI) studies depicting thepooled percentage of agreement with histopathology when assigningpostpolypectomy surveillance intervals based on combining real-time op-tical biopsy of colorectal polyps 5 mm or smaller, with histopathologicassessment of polyps larger than 5 mm.


of agreement with histopathology when assigning postpo-lypectomy surveillance intervals based on combining real-time FICE optical biopsy of colorectal polyps 5 mm orsmaller with histopathologic assessment of polyps largerthan 5 mm by using the MSTF postpolypectomy surveil-lance intervals. The reported degree of agreement ofFICE optical biopsy with histopathology in assigning post-polypectomy surveillance intervals based on these 2 studieswas 100% (95% CI, 91-100) and 97% (95% CI, 89 -100).

Confocal laser endomicroscopyRelatively few CLE studies have assessed this technology

in achieving PIVI thresholds. Several studies were per-formed by using postprocedure “offline” evaluation ofCLE images,19,38,41-44 and these were excluded from thismeta-analysis, which is limited to real-time studies evalu-ating polyps in situ. Six studies, which included 5 pub-lished studies45-49 and 1 abstract,50 reported real-timeclassification of CLE images in situ. Two of these studiescompared offline image assessment with real-time diag-nosis.47,49 Half of the studies used endoscopy-basedCLE,45,46,48 and half used probe-based CLE.47,49,50 Twostudies included chromoendoscopy as a means to targetareas for inspection with CLE.45,48 One study45 reportedthe correlation of CLE with all histology including bothpolypoid lesions and random sites assessed every 10 cmthroughout the colon, whereas other studies specificallytargeted polypoid lesions encountered during standard co-lonoscopy to assess the correlation of CLE with histol-ogy.46,47,49,50 This difference in the denominator where 1study includes endoscopically apparent normal areas can


greatly affect the performance of the diagnostic parame-ters, including NPV. Although most studies used fluores-cein sodium as the primary fluorophore, 2 studies alsoused acriflavin, a contrast agent that also stains the nu-cleus.45,48 One study compared the images of fluoresceinsodium versus acriflavin.45 The other study48 used acrifla-vin for nuclear staining, in conjunction with fluorescein so-dium; this study attempted to classify the differencebetween low-grade and high-grade dysplasia in suspect

www.giejournal.org


Subgroup by Practice Setting

Group by

Academic


MeanLowerlimit

Upperlimit

90% Agreement

Paggi 2012

Kuiper 2012

Ladabaum2013

Ignjatovic 2009

Rex 2009

Gupta 2012

Coe 2012

Repici 2013

Singh 2013

Wallace-1 2014

Wallace-2 2014

No

No

No

No

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Random

Random

85.28

81.48

79.90

82.4595.00

94.1286.10

70.0092.00

96.5595.00

94.0090.56

80.69 89.87

90.52

87.29

90.4598.94

96.9789.29

74.6895.98

98.9796.99

96.9994.95

72.44

72.51

74.4591.06

91.2782.91

65.3288.02

94.1393.01

91.0186.17

Figure 7. Forest plot of narrow-band imaging studies depicting the pooled percentage of agreement with histopathology when assigning postpolypec-tomy surveillance intervals based on combining real-time optical biopsy of colorectal polyps 5 mm or smaller with histopathologic assessment of polypslarger than 5 mm stratified by practice setting (academic vs community).

Subgroup by Endoscopist Expertise

Group byExpert


90% Agreement

Ignjatovic 2009Kuiper2012Coe2012Ladabaum2013

Rex 2009Gupta 2012Paggi 2012Repici 2013Singh 2013Wallace-1 2014Wallace-2 2014

No

No

NoNo

No

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Random

Random

MeanLowerlimit

Upperlimit

95.0081.4870.0079.9081.8794.1286.1085.2892.0096.5595.0094.0091.99

91.0672.4465.3272.5175.5091.2782.9180.6988.0294.1393.0191.0187.62

98.9490.5274.6887.2988.2496.9789.2989.8795.9898.9796.9996.9996.35

Figure 8. Forest plot of narrow-band imaging studies depicting the pooled percentage of agreement with histopathology when assigning postpolypec-tomy surveillance intervals based on combining real-time optical biopsy of colorectal polyps 5 mm or smaller with histopathologic assessment of polypslarger than 5 mm stratified by operator expertise (expert vs novice).


lesions compared with histopathology. Among the CLEstudies that classified polyps as adenomatous polyps inreal time, the NPVs for determining adenomatous histol-ogy were 79% (154 polyps),49 88% (107 polyps),50 92%(115 polyps),46 and 100% (32 polyps).47 Notably, the

www.giejournal.org

smallest study yielded the highest NPV, whereas the largerstudies had more modest NPV values. Given the variabilitypresent among these few studies with real-time polypcharacterization with CLE, a meta-analysis was notperformed.



Subgroup by Confidence Level

Group byHigh Confidence

Study name Statistics of each study

Random

NoNoNo

YesYesYesYesYesYes

Yes

YesYesYesRandom

Gupta 2012Coe 2012

Ignjatovic 2009Rex 2009Paggi 2012Kuiper 2012Repici 2013Singh 2013Ladabaum2013Wallace-1 2014Wallace-2 2014

MeanLower limit

Upper limit 90% Agreement

86.10 82.91 89.2970.0078.5395.0094.1285.2881.4892.0096.5579.9095.0094.0091.38

65.3271.5491.0691.2780.6972.4488.0294.1372.5193.0191.0188.02

74.6885.5298.9496.9789.8790.5295.9898.9787.2996.9996.9994.74

Figure 9. Forest plot of narrow-band imaging studies depicting the pooled percentage of agreement with histopathology when assigning postpolypec-tomy surveillance intervals based on combining real-time optical biopsy of colorectal polyps 5 mm or smaller with histopathologic assessment of polypslarger than 5 mm stratified by high confidence interpretation of the optical biopsy.

NPV for i-SCAN Optical Biopsy

Group byExpert


NPV 90%

Random

Random

No

NoNoNoNo

Yes

YesYesYes

Yes

Hong_1_2012Hong_2_2012Chan 2012Schachschal 2014

Hoffman_1_2010Hoffman_2_2010Lee 2011Pigo 2013

MeanLower limit

Upper limit

67.00

70.0069.0072.3197.0096.5094.7493.0096.16

76.2058.7171.0857.15

68.5961.03

94.5293.6590.7286.2694.39

75.2981.3282.8576.9776.0499.4899.3598.7699.7497.93

Figure 10. Forest plot of studies evaluating the narrow-band imaging (NBI) for i-SCAN–assisted optical biopsy for predicting adenomatous polyp histol-ogy of small/diminutive colorectal polyps stratified by operator expertise (expert vs novice).


DISCUSSION

The PIVI document on real-time endoscopic assessmentof the histology of diminutive colorectal polyps5 wascreated by the ASGE to promote and facilitate a potentialparadigm shift in the management of diminutive colorectalpolyps by using optical biopsy with endoscopic technolo-gies rather than histopathology for polyp characterizationand for decision making regarding polyp management aswell as assigning surveillance intervals. The PIVI document


established performance thresholds that needed to be metbefore widespread adoption of such technologies to mini-mize the risks of misclassification of polyps. If and whenfeasible, based on evolving endoscopic technology andincreasing endoscopist experience with these technolo-gies, such an approach would be more cost-effectivebecause of the cost savings associated with avoiding resec-tion of diminutive rectosigmoid hyperplastic polyps andpathology evaluation of resected diminutive adenomatouspolyps.

www.giejournal.org


NPV for FICE Optical Biopsy

Group byMagnification

Study name Statistics for each studyNPV 90%

Random

Random

NoNoNoNoNoYesYesYesYesYes

Pohl 2008Togashi 2009Buchner 2010Longcroft 2011

Dos Santos 2010Kim 2011Longcroft 2012Dos Santos 2012

MeanLower limit

Upper limit

77.0076.0050.0078.0073.9892.0083.0084.0079.0085.08

67.0661.5931.3072.0566.6985.0879.2175.1166.1778.93

86.9490.4168.7083.9581.2898.9286.7992.8991.8391.22

Figure 11. Forest plot of studies evaluating the negative predictive value (NPV) for Fujinon Intelligent Color Enhancement (FICE)–assisted optical biopsyfor predicting adenomatous polyp histology of small/diminutive colorectal polyps stratified by concurrent use of magnification.


Simulation Markov modeling has shown that a “resect-and-discard” strategy for diminutive polyps detected byscreening colonoscopy resulted in a substantial economicbenefit with a cost savings of $25 per person screened,which, projected to the U.S. population, would result inundiscounted annual savings of $33 million.51 Adding thecost savings from a “diagnose-and-leave” strategy, in whichthe cost of an endoscopic polypectomy is approximately$179 per person, would translate to a huge cost savingsto the U.S. health care system, estimated at more than$1 billion per year, without much impact on colonoscopyefficacy. In addition to the cost savings, such an approachwould avoid potential adverse events associated with un-necessary polypectomy of diminutive hyperplastic rectosig-moid polyps and would also be more efficient because inmany cases it would allow for immediate postprocedureassignment and communication of colonoscopy surveil-lance intervals to patients and referring physicians.

For a “diagnose-and-leave” strategy for diminutive recto-sigmoid polyps predicted to be nonneoplastic based onoptical biopsy, the PIVI recommends that endoscopic diag-nosis should provide a 90% or higher NPV for adenoma-tous histology when used with high confidence. For NBI,our meta-analysis indicated a pooled NPV of 91% usingthe random-effects model. Subgroup analyses indicatedthat NPVs were marginally higher in academic settingscompared with community settings, although this wasnot statistically significant. In addition, assessments madeby endoscopists experienced in optical biopsy polyp char-acterization, as well as optical biopsy assessments madewith high confidence, were associated with higher NPVs(93%). Our meta-analysis therefore indicates that opticalbiopsy technology by using NBI can meet this PIVIthreshold and supports a “diagnose-and-leave” strategyfor diminutive predicted nonneoplastic polyps in the recto-sigmoid colon.

www.giejournal.org

For a “resect-and-discard” (without pathology assess-ment) strategy for adenomas 5 mm or smaller, the PIVI rec-ommends that endoscopic characterization of polyphistology by optical biopsy (when used with high confi-dence), when combined with the histopathologic assess-ment of polyps larger than 5 mm, should provide a 90%or higher agreement in assignment of postpolypectomysurveillance intervals compared with decisions based onpathology assessment of all identified polyps. Subgroupanalyses of NBI studies indicated that the agreement inassignment of postpolypectomy surveillance intervals ex-ceeded 90% or higher in academic settings (90%), withexperienced endoscopists (92%) and when optical biopsyassessments were made with high confidence (91%). Thehighest agreement in assignment of postpolypectomy sur-veillance intervals was achieved with experienced endo-scopists making optical biopsy assessments with highconfidence (93%). Our meta-analysis therefore indicatesthat optical biopsy technology by using NBI can meetthis PIVI threshold and supports a “resect-and-discard”strategy for colorectal adenomas 5 mm or smaller.

In addition to NBI, our study included meta-analyses ofadditional advanced imaging technologies includingi-SCAN and FICE. Although our analyses indicate that, aswith NBI, the PIVI thresholds can potentially be met bythese technologies, in particular with expert operators,these technologies have not been as frequently studiedas NBI, and further studies need to be performed evalu-ating their use in characterizing diminutive colorectalpolyps. Studies using CLE demonstrate potential in thecharacterization of polyps. However, the majority of thesestudies were performed with postprocedure offline assess-ments. Additional CLE studies with real-time use of thistechnology will be beneficial in evaluating the ability ofthis technology to meet thresholds set by the PIVIdocument.




This meta-analysis was conducted only of studies withreal-time endoscopic assessment of polyps. Previousmeta-analyses, such as the study by Wanders et al,52

included real-time as well as postprocedure offline analysesto provide an overview of the potential of advanced imag-ing technology. However, because we were evaluating aparadigm that supports real-time decision making, welimited our analysis to studies performing real-time insitu evaluation.

The PIVI document also specifies that assessments ofpolyp histology using endoscopic technology be madewith high confidence. Not all studies have reported highconfidence assessments, but in subgroup analyses of thosethat did, we were able to demonstrate improved predictionof histology and reinforce the specification detailed in thePIVI that actionable assessments should be made with highconfidence. Our findings are also congruent with anotherrecent meta-analysis which indicated that when endo-scopic assessments of polyp histology are made withhigh confidence, the NPV for adenomatous histology andagreement with surveillance intervals based on histopa-thology alone exceed 90%.53

One limitation of our meta-analyses is the high degreeof heterogeneity among included studies. We correctedfor this by performing subgroup analyses to adjust forsome of the confounders contributing to this heterogene-ity. Another limitation is the lack of differentiation betweensmall and diminutive polyps in some of the includedstudies. Finally, this meta-analysis was limited to real-timepolyp assessments and therefore excluded studies withpostprocedure analyses.

As confirmed by our meta-analyses, expertise in inter-preting optical biopsies is a critical factor in optimizingtheir performance and meeting PIVI thresholds. This un-derscores the crucial need to train endoscopists in usingadvanced imaging technologies. Studies indicate that ac-curate interpretation of optical biopsies follows alearning curve, and this learning curve can be achievedrapidly in both academic and community practice set-tings.54-59

This ASGE Technology Committee systematic reviewand meta-analysis therefore confirms that the thresholdsestablished by the ASGE PIVI for the real-time endoscopicassessment of the histology of diminutive polyps havebeen met, at least for NBI, with endoscopists who are ex-perts in using these advanced imaging technologies andwhen assessments are made with high confidence. TheASGE Technology Committee therefore endorses the useof NBI for both the “diagnose-and-leave” strategy fordiminutive rectosigmoid hyperplastic polyps and the“resect-and-discard” strategy for diminutive adenomatouspolyps by endoscopists trained in using this technologyfor polyp characterization, making assessments with highconfidence.

Our analyses indicate that the PIVI thresholds canpotentially also be met by other advanced imaging technol-


ogies including i-SCAN, FICE, and CLE. However, given thelimited data on the use of these technologies in the char-acterization of diminutive colorectal polyps, further studiesneed to be performed.

Both of the PIVI-defined strategies for endoscopicassessment of the histology of diminutive colorectal polypsperturb the status quo and are likely to cause some trepi-dation among stakeholders including patients, endoscop-ists, and pathologists. The “diagnose-and-leave” strategycan be expected to cause more concern than the “resect-and-discard” strategy for both endoscopists and patients.Patients may have concerns regarding the accuracy ofpolyp characterization, particularly if polyps have beenleft unresected. Some endoscopists may have concernsregarding the increased unreimbursed work and responsi-bility, together with litigation risks they may face while tak-ing on a role traditionally performed by pathologists.Finally, some pathologists may be concerned regardingdisruptive endoscopic imaging technologies that mayaffect the scope of their work.

Future challenges therefore remain before the wide-spread implementation of these strategies in clinical prac-tice, including standardization of polyp classificationsystems based on advanced endoscopic imaging endo-scopic technologies, establishing standards of practice forthe use of these technologies in performing optical biopsy,developing training and credentialing programs, and estab-lishing quality metrics to help develop quality assuranceprograms. The ASGE Standards of Practice, Training,Educational Products, and Quality Assurance in Endoscopycommittees will address all of these issues to promote andfacilitate widespread adoption and implementation ofthese PIVI strategies.

DISCLOSURES

The following authors disclosed financial relationshipsrelevant to this article: Dr Banerjee has received researchfunding from Pentax. Dr Konda has received honorariafrom Manua Kea Technologies and grant support fromOlympus. Dr Wallace has received research fundingfrom Olympus. Dr Rex has received research supportand is a consultant for Olympus. All other authors dis-closed no financial relationships relevant to this article.

Abbreviations: ASGE, American Society for Gastrointestinal Endoscopy;CI, confidence interval; CLE, confocal laser endoscopy; FICE, FujinonIntelligent Color Enhancement; MSTF, Multi-Society Task Force; NBI,narrow-band imaging; NPV, negative predictive value; PIVI,Preservation and Incorporation of Valuable endoscopic Innovations.

REFERENCES

1. Gupta N, Bansal A, Rao D, et al. Prevalence of advanced histologicalfeatures in diminutive and small colon polyps. Gastrointest Endosc2012;75:1022-30.

www.giejournal.org

http://refhub.elsevier.com/S0016-5107(14)02592-9/sref1





2. O'Brien MJ, Winawer SJ, Zauber AG, et al. The National Polyp Study. Pa-tient and polyp characteristics associated with high-grade dysplasia incolorectal adenomas. Gastroenterology 1990;98:371-9.

3. Winawer SJ, Zauber AG, Ho MN, et al. Prevention of colorectal cancerby colonoscopic polypectomy. The National Polyp Study Workgroup.N Engl J Med 1993;329:1977-81.

4. Hussain ZH, Pohl H. Ancillary imaging techniques and adenoma detec-tion. Gastroenterol Clin North Am 2013;42:547-65.

5. Rex DK, Kahi C, O'Brien M, et al. The American Society for Gastrointes-tinal Endoscopy PIVI (Preservation and Incorporation of Valuable Endo-scopic Innovations) on real-time endoscopic assessment of thehistology of diminutive colorectal polyps. Gastrointest Endosc2011;73:419-22.

6. Schoenfeld P, Cook D, Hamilton F, et al. An evidence-based approachto gastroenterology therapy. Evidence-Based Gastroenterology Steer-ing Group. Gastroenterology 1998;114:1318-25.

7. Repici A, Hassan C, Radaelli F, et al. Accuracy of narrow-band imagingin predicting colonoscopy surveillance intervals and histology of distaldiminutive polyps: results from a multicenter, prospective trial. Gastro-intest Endosc 2013;78:106-14.

8. Wallace MB, Crook JE, Coe S, et al. Accuracy of in vivo colorectal polypdiscrimination by using dual-focus high-definition narrow-band imag-ing colonoscopy. Gastrointest Endosc 2014;80:1072-87.

9. Singh R, Jayanna M, Navadgi S, et al. Narrow-band imaging with dualfocus magnification in differentiating colorectal neoplasia. Dig Endosc2013;25(Suppl 2):16-20.

10. Ladabaum U, Fioritto A, Mitani A, et al. Real-time optical biopsy of co-lon polyps with narrow band imaging in community practice does notyet meet key thresholds for clinical decisions. Gastroenterology2013;144:81-91.

11. Gupta N, Bansal A, Rao D, et al. Accuracy of in vivo optical diagnosis ofcolon polyp histology by narrow-band imaging in predicting colonos-copy surveillance intervals. Gastrointest Endosc 2012;75:494-502.

12. Ignjatovic A, East JE, Suzuki N, et al. Optical diagnosis of small colo-rectal polyps at routine colonoscopy (Detect InSpect ChAracteriseResect and Discard; DISCARD trial): a prospective cohort study. LancetOncol 2009;10:1171-8.

13. East JE, Suzuki N, Bassett P, et al. Narrow band imaging with magnifi-cation for the characterization of small and diminutive colonic polyps:pit pattern and vascular pattern intensity. Endoscopy 2008;40:811-7.

14. Lee CK, Lee S-H, Hwangbo Y. Narrow-band imaging versus I-SCAN forthe real-time histological prediction of diminutive colonic polyps: aprospective comparative study by using the simple unified endoscopicclassification. Gastrointest Endosc 2011;74:603-9.

15. Paggi S, Rondonotti E, Amato A, et al. Resect and discard strategy inclinical practice: a prospective cohort study. Endoscopy 2012;44:899-904.

16. Hewett DG, Huffman ME, Rex DK. Leaving distal colorectal hyperplasticpolyps in place can be achieved with high accuracy by using narrow-band imaging: an observational study. Gastrointest Endosc 2012;76:374-80.

17. Hewett DG, Kaltenbach T, Sano Y, et al. Validation of a simple classifi-cation system for endoscopic diagnosis of small colorectal polyps us-ing narrow-band imaging. Gastroenterology 2012;143:599-607.e1.

18. Rex DK. Narrow-band imaging without optical magnification for histo-logic analysis of colorectal polyps. Gastroenterology 2009;136:1174-81.

19. Shahid MW, Buchner AM, Heckman MG, et al. Diagnostic accuracy ofprobe-based confocal laser endomicroscopy and narrow band imag-ing for small colorectal polyps: a feasibility study. Am J Gastroenterol2012;107:231-9.

20. Kuiper T, Marsman WA, Jansen JM, et al. Accuracy for optical diagnosisof small colorectal polyps in nonacademic settings. Clin GastroenterolHepatol 2012;10:1016-20; quiz e79.

21. Henry ZH, Yeaton P, Shami VM, et al. Meshed capillary vessels foundon narrow-band imaging without optical magnification effectivelyidentifies colorectal neoplasia: a North American validation of the Jap-anese experience. Gastrointest Endosc 2010;72:118-26.

www.giejournal.org

22. Rogart JN, Jain D, Siddiqui UD, et al. Narrow-band imaging withouthigh magnification to differentiate polyps during real-time colonos-copy: improvement with experience. Gastrointest Endosc 2008;68:1136-45.

23. Sakamoto T, Matsuda T, Aoki T, et al. Time saving with narrow-band imaging for distinguishing between neoplastic and non-neoplastic small colorectal lesions. J Gastroenterol Hepatol2012;27:351-5.

24. Sano Y, Ikematsu H, Fu KI, et al. Meshed capillary vessels by use ofnarrow-band imaging for differential diagnosis of small colorectalpolyps. Gastrointest Endosc 2009;69:278-83.

25. van den Broek FJC, Fockens P, Van Eeden S, et al. Clinical evaluation ofendoscopic trimodal imaging for the detection and differentiation ofcolonic polyps. Clin Gastroenterol Hepatol 2009;7:288-95.

26. Coe SG, Thomas C, Crook J, et al. Colorectal surveillance intervalassignment based on in vivo prediction of polyp histology: impactof endoscopic quality improvement program. Gastrointest Endosc2012;76:118-25.e1.

27. Schachschal G, Mayr M, Treszl A, et al. Endoscopic versus histologicalcharacterisation of polyps during screening colonoscopy. Gut2014;63:458-65.

28. Pigo F, Bertani H, Manno M, et al. i-SCAN high-definition white lightendoscopy and colorectal polyps: prediction of histology, interob-server and intraobserver agreement. Int J Colorectal Dis 2013;28:399-406.

29. Hong SN, Choe WH, Lee JH, et al. Prospective, randomized, back-to-back trial evaluating the usefulness of i-SCAN in screening colonos-copy. Gastrointest Endosc 2012;75:1011-1021 e2.

30. Chan JL, Lin L, Feiler M, Wolf AI, et al. Comparative effectiveness ofi-SCAN and high-definition white light characterizing small colonicpolyps. World J Gastroenterol 2012;18:5905-11.

31. Hoffman A, Sar F, Goetz M, et al. High definition colonoscopy com-bined with i-SCAN is superior in the detection of colorectal neoplasiascompared with standard video colonoscopy: a prospective random-ized controlled trial. Endoscopy 2010;42:827-33.

32. Hoffman A, Kagel C, Goetz M, et al. Recognition and characterization ofsmall colonic neoplasia with high-definition colonoscopy using i-SCANis as precise as chromoendoscopy. Dig Liver Dis 2010;42:45-50.

33. Longcroft-Wheaton G, Brown J, Cowlishaw D, et al. High-definition vs.standard-definition colonoscopy in the characterization of smallcolonic polyps: results from a randomized trial. Endoscopy 2012;44:905-10.

34. Dos Santos CE, Malaman D, Lopes CV, et al. Digital chromoendoscopyfor diagnosis of diminutive colorectal lesions. Diagn Ther Endosc2012;2012:279521.

35. Longcroft-Wheaton GR, Higgins B, Bhandari P. Flexible spectral imag-ing color enhancement and indigo carmine in neoplasia diagnosis dur-ing colonoscopy: a large prospective UK series. Eur J GastroenterolHepatol 2011;23:903-11.

36. Kim YS, Kim D, Chung SJ, et al. Differentiating small polyp histologiesusing real-time screening colonoscopy with Fuji Intelligent ColorEnhancement. Clin Gastroenterol Hepatol 2011;9:744-749 e1.

37. dos Santos CE, Lima JC, Lopes CV, et al. Computerized virtual chro-moendoscopy versus indigo carmine chromoendoscopy combinedwith magnification for diagnosis of small colorectal lesions: a random-ized and prospective study. Eur J Gastroenterol Hepatol 2010;22:1364-71.

38. Buchner AM, Shahid MW, Heckman MG, et al. Comparison of probe-based confocal laser endomicroscopy with virtual chromoendoscopyfor classification of colon polyps. Gastroenterology 2010;138:834-42.

39. Pohl J, Lotterer E, Balzer C, et al. Computed virtual chromoendoscopyversus standard colonoscopy with targeted indigocarmine chromo-scopy: a randomised multicentre trial. Gut 2009;58:73-8.

40. Togashi K, Osawa H, Koinuma K, et al. A comparison of conventionalendoscopy, chromoendoscopy, and the optimal-band imaging systemfor the differentiation of neoplastic and non-neoplastic colonic polyps.Gastrointest Endosc 2009;69:734-41.










































































































































41. Kuiper T, van den Broek FJ, van Eeden S, et al. Feasibility and accuracyof confocal endomicroscopy in comparison with narrow-band imagingand chromoendoscopy for the differentiation of colorectal lesions. AmJ Gastroenterol 2012;107:543-50.

42. Buchner AM, Gomez V, Heckman MG, et al. The learning curve ofin vivo probe-based confocal laser endomicroscopy for prediction ofcolorectal neoplasia. Gastrointest Endosc 2011;73:556-60.

43. Andre B, Vercauteren T, Buchner AM, et al. Software for automatedclassification of probe-based confocal laser endomicroscopy videosof colorectal polyps. World J Gastroenterol 2012;18:5560-9.

44. Gomez V, Buchner AM, Dekker E, et al. Interobserver agreement andaccuracy among international experts with probe-based confocal laserendomicroscopy in predicting colorectal neoplasia. Endoscopy2010;42:286-91.

45. Kiesslich R, Burg J, Vieth M, et al. Confocal laser endoscopy for diag-nosing intraepithelial neoplasias and colorectal cancer in vivo. Gastro-enterology 2004;127:706-13.

46. Xie XJ, Li CQ, Zuo XL, et al. Differentiation of colonic polyps by confocallaser endomicroscopy. Endoscopy 2011;43:87-93.

47. De Palma GD, Staibano S, Siciliano S, et al. In vivo characterisation ofsuperficial colorectal neoplastic lesions with high-resolution probe-based confocal laser endomicroscopy in combination with video-mosaicing: a feasibility study to enhance routine endoscopy. Dig LiverDis 2010;42:791-7.

48. Sanduleanu S, Driessen A, Gomez-Garcia E, et al. In vivo diagnosis andclassification of colorectal neoplasia by chromoendoscopy-guidedconfocal laser endomicroscopy. Clin Gastroenterol Hepatol 2010;8:371-8.

49. Shahid MW, Buchner AM, Raimondo M, et al. Accuracy of real-time vs.blinded offline diagnosis of neoplastic colorectal polyps using probe-based confocal laser endomicroscopy: a pilot study. Endoscopy2012;44:343-8.

50. Singson Z, Hashemzadeh M, Jamal MM. Utilization of probe-basedconfocal laser endomicroscopy in a resect and discard approachto small colon polyps [abstract]. Gastrointest Endosc 2012;75:AB224.

51. Hassan C, Pickhardt PJ, Rex DK. A resect and discard strategy wouldimprove cost-effectiveness of colorectal cancer screening. Clin Gastro-enterol Hepatol 2010;8:865-9.

52. Wanders LK, East JE, Uitentuis SE, et al. Diagnostic performance of nar-rowed spectrum endoscopy, autofluorescence imaging, and confocallaser endomicroscopy for optical diagnosis of colonic polyps: ameta-analysis. Lancet Oncol 2013;14:1337-47.

53. McGill SK, Evangelou E, Ioannidis JPA, et al. Narrow band imaging todifferentiate neoplastic and non-neoplastic colorectal polyps in real


time: a meta-analysis of diagnostic operating characteristics. Gut2013;62:1704-13.

54. Neumann H, Vieth M, Fry LC, et al. Learning curve of virtual chromoen-doscopy for the prediction of hyperplastic and adenomatous colo-rectal lesions: a prospective 2-center study. Gastrointest Endosc2013;78:115-20.

55. Rastogi A, Rao DS, Gupta N, et al. Impact of a computer-based teachingmodule on characterization of diminutive colon polyps by usingnarrow-band imaging by non-experts in academic and communitypractice: a video-based study. Gastrointest Endosc 2014;79:390-8.

56. Bouwens MW, de Ridder R, Masclee AA, et al. Optical diagnosis of colo-rectal polyps using high-definition i-SCAN: an educational experience.World J Gastroenterol 2013;19:4334-43.

57. Kuiper T, Kiesslich R, Ponsioen C, et al. The learning curve, accuracy,and interobserver agreement of endoscope-based confocal laser en-domicroscopy for the differentiation of colorectal lesions. GastrointestEndosc 2012;75:1211-7.

58. Raghavendra M, Hewett DG, Rex DK. Differentiating adenomas fromhyperplastic colorectal polyps: narrow-band imaging can be learnedin 20 minutes. Gastrointest Endosc 2010;72:572-6.

59. Higashi R, Uraoka T, Kato J, et al. Diagnostic accuracy of narrow-bandimaging and pit pattern analysis significantly improved for less-experienced endoscopists after an expanded training program. Gastro-intest Endosc 2010;72:127-35.

Prepared by:ASGE TECHNOLOGY COMMITTEEBarham K. Abu Dayyeh, MD, MPHNirav Thosani, MDVani Konda, MDMichael B. Wallace, MD, MPH (invited expert, previous Committee Co-Chair)Douglas K. Rex, MD (invited expert)Shailendra S. Chauhan, MDJoo Ha Hwang, MDSri Komanduri, MDMichael Manfredi, MD, NASPGAN RepresentativeJohn T. Maple, DOFaris M. Murad, MDUzma D. Siddiqui, MDSubhas Banerjee, MD, Committee Chair

This document was developed by the ASGE Technology Committee. Thisdocument was reviewed and approved by the governing board of theAmerican Society for Gastrointestinal Endoscopy (ASGE).

www.giejournal.org







































































Appendix 1. Publication bias funnel plot for narrow-band imagingstudies

Appendix 2. Publication bias funnel plot for i-SCAN studies

Appendix 3. Publication bias funnel plot for FICE studies

www.giejournal.org Volume 81, No. 3 : 2015 GASTROINTESTINAL ENDOSCOPY 516.e1



Documents

ASGE Technology Committee systematic review and meta