Article Text
PDF
Abstract
Background Diminutive (≤5 mm) colorectal polyps are common, and overwhelmingly benign. Routinely, after polypectomy, they are examined pathologically to determine the surveillance intervals. Advances in equipment and techniques, such as narrow-band imaging (NBI) colonoscopy, now permit reliable real-time optical diagnosis.
Methods We conducted a randomised single-masked study involving three institutions to determine whether optical diagnosis of diminutive colorectal polyps meets clinical practice standards and reduces the need for histopathology. We randomly assigned eligible patients undergoing routine high-definition colonoscopy to optical diagnosis using near focus versus standard view, using computer-generated block sequence. By validated criteria, we rendered an optical diagnosis and a confidence level (high vs low) for all polyps, using NBI. Our primary endpoint was the number of accurate high-confidence optical diagnoses compared with central blinded pathology in the two groups. We analysed data using intention to treat.
Findings We enrolled 558 subjects, and randomly assigned 281 to near focus and 277 to standard view optical diagnosis. We detected 1309 predominantly diminutive (74.5%) and neoplastic (60.0%) polyps. Endoscopists were significantly more likely, OR 2.2 (95% CI 1.6 to 3.0, p<0.0001), to make a high-confidence optical diagnosis with near focus (85.1%) than standard (72.6%) view. High-confidence diagnoses had 96.4% and 92.0% negative predictive value, respectively. Of all polyps, 75.3% (95% CI71.3% to 78.9%) had a high-confidence accurate prediction using near focus, compared with 63.1% (95% CI 58.5% to 67.6%) using standard view. Optical versus histopathological diagnosis showed excellent agreement between the surveillance intervals, 93.5% in near focus and 92.2% in standard view. The median diagnosis time was 14 s.
Conclusions Real-time optical diagnosis using NBI colonoscopy may replace the pathology diagnosis for the majority of diminutive colorectal polyps. Using colonoscopy with near focus view increases the confidence level of the optical diagnosis. Optical diagnosis would be a paradigm shift in clinical practice of colonoscopy for colorectal cancer screening.
Trial registration number ClinicalTrials.gov Identifier: NCT01288833.
- COLONOSCOPY
- COLORECTAL NEOPLASIA
- COLORECTAL CANCER SCREENING
- COLORECTAL PATHOLOGY
- IMAGING
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Colonoscopy with polypectomy is effective at preventing colorectal cancer deaths, with the majority of polyps detected and removed during colonoscopy being diminutive in size and benign.
Current practice, after polypectomy, is routine pathologic analysis of all colorectal polyps in order to provide a diagnosis to determine the surveillance interval.
Neoplastic polyps have distinct colour, vessel and surface features as the pattern and size of microvessels in the mucosa and submucosal change when tissue becomes neoplastic. The Narrow-Band Imaging (NBI) International Colorectal Endoscopic (NICE) classification has been validated for the real-time endoscopic characterisation of colorectal polyps.
Meta-analyses from academic centres on optical diagnosis of colorectal polyps have shown similar diagnostic characteristics to histopathology—93% overall concordance of surveillance intervals and ≥90% negative predictive value for small polyps in the rectosigmoid colon.
What are the new findings?
Currently available colonoscopes are now equipped with dual focus imaging capabilities for both magnified near focus (65×) and standard (30×) viewing using a push-button.
Using the near focus view for optical diagnosis increases the confidence level of the endoscopist, as they were significantly more likely to make a high-confidence optical diagnosis with near focus (85.1%) than standard (72.6%) view, p<0.0001.
This is the first prospective randomised multicentre trial in the USA to show that optical diagnosis of diminutive colorectal polyps by NBI with near focus view, using validated classification criteria, can surpass diagnostic thresholds set forth in the clinical guidelines.
All five endoscopists met the thresholds with overall 93.5% agreement between the surveillance intervals of optical versus histopathological diagnosis, and 96.4% negative predictive value for diminutive polyps in the rectosigmoid colon.
How might it impact on clinical practice in the foreseeable future?
Optical diagnosis of colorectal polyps could improve cost-effectiveness of screening colonoscopy with potential cost reduction of pathology, improvements in efficiency related to the follow-up care and reduction in patient anxiety associated with the delay in the pathology results.
Optical diagnosis cannot be applied to every polyp, and thus, the use of confidence levels is important in the strategy. If a polyp lacks features associated with confident endoscopic assignment of histology, the endoscopist could still resect it and submit for pathologic assessment.
Introduction
Most colorectal polyps are diminutive (≤5 mm).1 At that size, they are overwhelmingly hyperplastic or adenomatous, and rarely harbour high-grade dysplasia, cancer (0.06%)2 or sessile serrated adenoma/polyp.3 Yet they are routinely removed4 and submitted for pathologic analyses, primarily to determine the surveillance colonoscopy interval.5 A strategy that relies on real-time endoscopic diagnosis of diminutive polyps without the histopathological assessment could improve the efficiency of colonoscopy for colorectal cancer screening6 and save millions at a time when colonoscopy costs continue to rise.7 ,8 The technologies, techniques and standards of practice to implement the optical diagnosis of diminutive colorectal polyps have recently been developed. Newer colonoscopes are equipped with an image enhancement feature, such as narrow-band imaging (NBI), i-Scan or Fujinon intelligent chromoendoscopy (FICE), which highlights mucosal microvascularity patterns that reflect the histology. An endoscopic classification of colorectal polyp histopathology has been validated.9 The standards for the practice of endoscopic diagnosis for diminutive colorectal polyps has been published.10
Real-time endoscopic diagnosis of colorectal polyp histology has been shown to be highly accurate using NBI.11 Though optical diagnosis of polyps has similar diagnostic operating characteristics as pathologic diagnosis and has the potential to shift the current practice paradigm, endoscopists have been slow to adapt the strategy into clinical practice. We consider the confidence that the endoscopist has in their endoscopic prediction of colorectal polyp histology to be an important factor. As such, we hypothesised that near focus, detailed viewing of the polyp pattern, using colonoscopes with push-button optical magnification, would improve the endoscopists’ confidence in the optical diagnosis, and thus, augment the implementation in clinical practice. We aimed to study the confidence level and accuracy of optical diagnosis versus histopathological diagnosis of diminutive colorectal polyps. We compared colonoscopes capable of near focus viewing (65×) to that of standard viewing (30×).
Methods
Study design
The Veterans Affairs Colorectal Lesion Interpretation and Diagnosis (VALID) study was a multicentre, single-blinded, randomised trial involving the Veterans Affairs Palo Alto, Kansas City and San Francisco Health Care Systems (ClinicalTrials.gov Identifier: NCT01288833). The institutional review boards approved the protocol. All patients provided written informed consent.
Patient selection, randomisation and masking
From March 2011 to May 2012, we consecutively recruited patients undergoing routine colonoscopy (table 1). We excluded patients from enrolment who were referred for polypectomy or had colitis, a personal or family history of polyposis or hereditary colorectal cancer syndrome, or coagulopathy/thrombocytopenia. A statistician centrally generated randomisation numbers that were stratified according to site and procedure indication, using the R package blockrand.12 ,13 Using sealed opaque envelopes, a coordinator randomly assigned individuals into one of two groups: dual-focus colonoscope (CFHQ190AL, EVIS Exera III), which includes a near focus (close view) and standard focus (standard view); or a standard-focus colonoscope (CFH180AL, EVIS Exera II), which includes only standard view. High-definition 26-inch liquid crystal display (LCD) monitors were used in both groups (OEV-261H). Subjects and pathologists were masked to the randomisation assignment. Endoscopists were unmasked due to the real-time visualisation.
Baseline characteristics
Optical diagnosis colonoscopy
Training
Five endoscopists participated. Before enrolment, all completed a standardised training in optical diagnosis of colorectal polyp histology using a Learning Management System, exceeding 90% accuracy. They used the NBI International Colorectal Endoscopic (NICE) classification.9
Procedure
All patients received split bowel preparation with polyethylene glycol-based electrolyte solution and/or magnesium citrate, consuming purgative on the previous evening and on the morning of the colonoscopy.
In both groups, the endoscopist performed initial examination of the colon using high definition white light colonoscopy with standard view. Once a polyp was detected, optical diagnosis was made using the NBI mode. In the dual-focus colonoscope arm, it was performed in near focus view mode (65X) and in the standard-focus colonoscope arm, in the standard view mode (30X). The endoscopist diagnosed the polyp histology as neoplastic (tubular adenoma, villous adenoma, high-grade dysplasia, cancer) or non-neoplastic (hyperplastic, sessile serrated adenoma/polyp3 or inflammatory), and provided a confidence level (high or low). High confidence was assigned if the polyp had one or more features of Type 2 (neoplasia) or Type 1 (non-neoplasia) in the NICE classification and no features associated with the other histology. Alternatively, low confidence was assigned. Endoscopists also assigned low confidence if the lesion showed features suggestive of a sessile serrated adenoma/polyp, as such lesions exhibit characteristics of non-neoplasia but have not been validated in the classification.
We estimated polyp size using a biopsy forceps or snare, location on anatomic landmarks, and shape on morphology.14 ,15 We digitally captured three images for each polyp. Following polypectomy, we immediately and separately submitted each polyp for pathologic processing and analysis. In cases of multiple non-neoplastic (≥2) polyps in the rectosigmoid, we obtained a representative sample. Using timestamps, the digital database measured insertion, withdrawal and polyp diagnosis, removal and retrieval times. The endoscopist provided a surveillance interval recommendation based upon the optical diagnoses using current guidelines.5 ,16 The study protocol, case report forms and images are available.
Histopathology analysis
A central gastrointestinal pathologist who was blinded to the endoscopic diagnosis, reviewed each specimen. Polyps were defined as adenoma or hyperplastic according to WHO criteria.17 A diagnosis of sessile serrated adenoma/polyp required architectural disturbances with serrations near the bases, proliferation from the bases of ≥3 contiguous crypts, and straight, narrow bases in fewer than half the crypts.
The central pathology diagnosis was compared with the initial local pathology diagnosis. A second blinded pathologist adjudicated any discrepancies between the diagnoses to determine the final diagnosis. We recut the tissue block for re-examination when a final pathologic diagnosis was discordant with a high-confidence optical diagnosis.
Study outcomes
The primary outcome measure was the proportion of accurate high-confidence optical diagnoses of neoplastic and non-neoplastic diminutive colorectal polyps using near focus compared with standard view.
A secondary objective was to compare the optical diagnostic characteristics including accuracy, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV), using the high-definition, NBI colonoscope with and without near focus view, based on final histopathology as the reference standard. We defined sensitivity as the proportion of neoplastic polyps that were endoscopically predicted to be neoplastic out of all neoplastic polyps diagnosed by histopathology, specificity as the proportion of non-neoplastic polyps that were predicted to be non-neoplastic out of all non-neoplastic polyps diagnosed by histopathology, PPV as the proportion of neoplastic polyps of those endoscopically predicted to be neoplastic, and NPV as the proportion of non-neoplastic polyps of those endoscopically predicted to be non-neoplastic.
Additional objectives were the agreement of surveillance interval recommendations based on optical diagnosis compared to those based on histopathology, colonoscopy quality indicators18 and adverse events.19
Statistical analysis
We calculated a sample size by the difference in proportions of accurate high-confidence optical diagnoses of colorectal polyps between the dual focus and the standard focus colonoscope groups. Optical diagnoses were defined as accurate if they matched the final histopathological diagnosis. We assumed 90% of polyps would be diagnosed with high confidence using the near focus view, and 80% using the standard view. Based on our earlier data,20 assuming a 97% caecal intubation rate, a 5% poor bowel preparation quality, and that 60% would have a colorectal lesion with a mean neoplasm of 0.85 per patient, we determined enrolment of 279 patients in each group to have a statistical power of 80% with a two-sided level of 0.05.
We performed intention-to-treat analyses using Stata/SE 12.1. We performed logistic regression using a conditional random-effects model clustered by a patient in order to account for a field effect of multiple polyps within a patient. The ORs with 95% CI served to describe the influence of the dual focus modality, which allowed for comparisons of high and low confidence proportions. We explored the impact of lesion size and morphology on the model, and assessed the endoscopist as a confounder. For univariate analysis, we used the t test or Wilcoxon rank sum test for comparison of normally distributed or non-parametric continuous predictors, respectively. We estimated diagnostic operating characteristics. We classified a lost (not retrieved) specimen as an incorrect optical diagnosis.
Role of funding source
This study was partially funded by Olympus Medical America. Neither investigators nor patients received financial compensation for the trial. The investigators designed the trial, gathered, analysed and maintained the data, and decided to submit the manuscript for publication. The first and senior authors drafted the manuscript, and all authors contributed to its revision. The sponsor had no role.
Results
Patients
We enrolled and randomly assigned 558 patients (figure 1) with well-balanced characteristics (table 1). Over three-quarters of the patients had at least one polyp (76.9%), and over half had at least a diminutive polyp (66.1%). Of the screening patients, 53.3% had at least one adenoma. Endoscopists diagnosed 1309 polyps, of which 975 (74.5%) were diminutive (table 2). None of the diminutive polyps (95% CI 0% to 0.38%) harboured high-grade dysplasia or cancer, and 13 (1.3%, 95% CI 0.71% to 2.3%) were sessile serrated adenoma/polyps.
Diminutive (≤5 mm) polyp characteristics (N=975)
Consort diagram. A total of 827 patients were assessed for eligibility. A total of 269 patients were excluded, mostly for not meeting inclusion criteria or for unavailable equipment at the time of the study procedure. No patients were lost to follow-up or discontinued the intervention, and 277 (near focus view) and 281 (standard view) were studied.
Optical diagnosis
Using the near focus view compared with the standard view, endoscopists made significantly more high-confidence optical diagnoses of diminutive polyps (figure 2), OR 2.2, 95% CI 1.6 to 3.0; p<0.0001, with 451 of 530 (85.1%, 95% CI 81.8% to 88.0%) in the near focus view group compared with 323 of 445 (72.6%, 95% CI 68.2 to 76.7) in the standard group. Of all diminutive polyps, 75.3% (95% CI 71.3% to 78.9%) had a high-confidence accurate prediction using near focus (399/530), compared with 63.1% (95% CI 58.5% to 67.6%) using standard view (281/445).
Diminutive colorectal polyps. Hyperplastic polyp: (A) Endoscopic image of a polyp in narrow-band imaging (NBI) near focus view shows same or lighter colour than the normal surrounding mucosa, isolated lacy vessels coursing across the lesion, and white spots of uniform size. (B) Histopathology image shows cytologically non-dysplastic mucosal lesion with serrations in the luminal portions of the glands and elongated, straight and narrow crypt bases. Adenomatous polyp: (C) Endoscopic image of the polyp in NBI near focus view shows browner colour than the normal surrounding mucosa, brown vessels surrounding white structures, and a surface structure with tubular and branched white structures surrounded by brown vessels. (D) Histopathology image shows low-grade dysplasia of the cells lining the surface and luminal portion of the glands from the 12 o'clock to 3 o'clock positions (normal mucosa is present from the 6 o'clock to 9 o'clock positions).
The optical diagnostic performance for diminutive polyps of the entire colorectum and by location is shown in table 3. Importantly, endoscopists achieved high performance when they had high confidence in their optical diagnosis. Using near focus view, the sensitivity was 98.2% (95% CI 95.5% to 99.3%) and NPV was 96.4% (95% CI 91.8% to 98.8%); for the standard view, the sensitivity was 95.2% (95% CI 90.8% to 97.6%) and NPV was 92.0% (95% CI 85.3% to 96.3%). Limited to rectosigmoid polyps, the NPV was 96.6% (95% CI 91.4% to 99.1%) using the near focus view, and 93.6% (95% CI 85.7% to 97.9%) using the standard view.
Diagnostic performance of optical diagnosis of diminutive colorectal polyps, made with high confidence (n=774)
Estimates of the diagnostic performance for individual endoscopists are shown in figure 3, stratified by confidence level. Using the near focus view, every endoscopist exceeded 90% sensitivity and NPV. The proportion of high-confidence predictions for diminutive polyps was consistently higher using the near focus view (range 81.7%–90.8%) compared with the standard view (range 59.5%–80.4%).
High-confidence and low-confidence diagnostic performance characteristics, by endoscopist using near focus view versus standard view colonoscopes. Using the near focus view, every endoscopist exceeded 90% sensitivity and NPV. Exact 95% CIs were calculated for each endoscopist, and large-sample 95% CIs were calculated for overall estimates, with the use of standard errors that allowed for estimation of variation among endoscopists. PPV, positive predictive value; NPV, negative predictive value.
Discrepancies between the local and centralised pathologic diagnoses occurred in 5.0% (95% CI 3.7% to 6.6%) of cases, in which a second blinded pathologist reviewed the slides to determine a final pathologic diagnosis. Compared to final pathology, the diagnostic performance of the original local pathology review showed sensitivity of 95.4%, 95% CI 93.3% to 96.9%, and NPV of 93.7%, 95% CI 90.9% to 95.9%. Furthermore, recutting and re-examination of the high-confidence optical diagnosis specimens that were discordant with the pathologic diagnosis (9.0%, 95% CI 7.3% to 11.0%, n=88/975) resulted in a change in the final pathologic diagnosis in 19.3% (95% CI 11.7% to 29.1%, n=17/88) of cases.
Surveillance interval agreement
At the end of the procedure, the optical diagnoses resulted in the correct surveillance interval recommendation in 92.8% (95% CI 90.4% to 94.8%) of the cohort (n=518/558), compared with the recommendation based on histopathological diagnosis. Both groups received correct surveillance recommendations, 93.5% (n=259/277) in the near focus view and 92.2% (n=259/281) in the standard view colonoscopy groups. Compared with surveillance interval recommendations based on the final pathology, the accuracy of optical diagnosis was similar to that of the initial pathology diagnosis.
The accuracy of surveillance interval recommendations increased to 96.6% (95% CI 94.7% to 97.9%) when it was based on combining the high-confidence optical diagnosis of diminutive colorectal polyps with the histopathology results of all other polyps (table 4). In the near focus view group, nine patients were instructed to return 2.9±1.3 years earlier. In the standard view group, seven had been told to return 2.4±1.1 years earlier; and three patients had delayed follow-up recommendations by 3.0±1.7 years.
Surveillance interval recommendation—agreement of optical diagnosis-based surveillance interval recommendations with pathology
The accuracy of the initial pathology-based surveillance interval recommendations compared to final pathology was 95.0% (95% CI 93.4% to 96.3%). Thirteen patients had a change in the interval following final pathology diagnosis—a delay in ten, and early follow-up in three.
Adverse events
There was no postpolypectomy bleeding, coagulation syndrome, or perforation. There was no optical misdiagnosis of advanced histology. Optical diagnosis did not significantly lengthen the procedure time: 15 s using near focus versus 12 s using standard view. The mean inspection time was 10 min.
Discussion
Our study suggests that optical diagnosis of diminutive colorectal polyps can be used in routine colonoscopies, using a standardised approach and validated criteria. We have demonstrated that accurate real-time diagnosis of polyp histopathology can be made with currently available endoscopic imaging, and that the majority of cases can be examined with high confidence. Among high-confidence optical diagnoses, the accuracy and NPVs were excellent. Indeed, high confidence in the optical diagnosis correlated with high diagnostic performance, serving as a clinical decision gauge. Colonoscopes equipped with near focus view facilitated a significantly greater percentage of high-confidence polyp predictions, and thus could broaden the implementation of the optical diagnosis strategy to routine clinical practice. Moreover, our findings show that patients can be informed of the polyp results at discharge, and that correct surveillance interval recommendations can be delivered, 92.8% (95% CI 90.4% to 94.8%). Importantly, even small improvements of significance can largely impact practice and economics given the high prevalence of colorectal polyps and the common performance of colonoscopies.
Our study findings highlight that diminutive polyps represent the bulk of polyps (74.5%) detected in colonoscopy and rarely harbour high-risk pathology (none had high-grade dysplasia or cancer; one had villous features). Current standard of practice, however, is to remove all polyps to prevent their growth and submit them for histopathology to determine the correct surveillance interval. The universal pathology of colorectal polyps results in significant resource usage for pathology processing and interpretation, follow-up of pathology results, and contacting patients and referring physicians with results and surveillance recommendations. As an alternative, the American Society for Gastrointestinal Endoscopy (ASGE) proposed, that for diminutive polyps to be resected and discarded without pathologic assessment, the endoscopic diagnosis should provide a ≥90% agreement in assigning surveillance interval compared with pathology-based diagnosis; ASGE also proposed that to leave diminutive rectosigmoid non-neoplastic polyps in situ without removal, the endoscopic diagnosis should provide ≥90% NPV.10 The recent European Society of Gastrointestinal Endoscopy guideline suggests that image-enhanced endoscopy with virtual (NBI, FICE, i-SCAN) and conventional chromoendoscopy can be used, under strictly controlled conditions, for real-time optical diagnosis of diminutive (≤5 mm) colorectal polyps to replace histopathological diagnosis.21 In our study, using validated scales and adequate photo documentation, the proposed clinical thresholds were exceeded: 93.5% (95% CI 89.9% to 96.1%) correct surveillance interval recommendations, and 96.6% (95% CI 91.4% to 99.1%) NPV.
This is the first prospective randomised multicentre trial in the USA to show that optical diagnosis of diminutive colorectal polyps by NBI with near focus view, using validated classification criteria, can surpass diagnostic thresholds set forth in the clinical guidelines. The accuracy of optical diagnosis of diminutive polyps, though not 100%, is equivalent to the accuracy of pathologic diagnosis in community practice. In our study, we found 5.0% (95% CI 3.7% to 6.6%) disagreement between the original and the final pathologic diagnoses. Others have reported of similar findings.22 ,23 In clinical practice, inaccuracies in histopathology may be even higher. Up to 6% of polyps, in fact, are not retrieved for histopathology. Additionally, the normal surrounding tissue in a polypectomy specimen of a diminutive lesion may be what is sectioned and interpreted. In this study, we performed recutting in 88 of the 975 diminutive polyp specimens, 9.0% (95% CI 7.3% to 11.0%), because the high-confidence optical diagnoses did not agree with the pathology diagnosis. Recuts of these blocks changed 19.3% (95% CI 11.7% to 29.1%) of the pathologic diagnosis, underscoring the overall limited evaluation of the gross polyp specimen in routine pathologic diagnosis.
A reduction in pathology specimens may improve the efficiency of the procedure.8 ,24 Our study confirms that an optical diagnosis strategy would decrease the number of specimens submitted for pathology, and enables us to relay findings to the patient on the procedure day in the vast majority cases. In our study, 241 bottles (grouped by colon segment) in the near focus view group could have been omitted from pathologic review based on the high confidence optical diagnosis, resulting in a direct pathology costs savings of US$167 per patient. Although we did not measure the indirect savings of reduced histopathology follow-up and notifications, they likely are substantial. Similarly, both direct communication to the patient and documentation in the report of the surveillance interval recommendation creates a reliable and consistent system. As opposed to the current system that states pending pathology, real-time relay of information in the report and to the patient avoids the potential loss of follow-up or unknown surveillance interval recommendations if the patient leaves the healthcare system. A formal cost-effectiveness analysis is underway.25
Serrated polyps, previously collectively defined as hyperplastic, are now distinctly subclassified as hyperplastic, sessile serrated adenoma/polyp and traditional serrated adenoma. The distinction stems from evidence indicating a malignant potential–particularly those of larger size, sessile serrated adenoma/polyp histology, cytologic dysplasia, and located in the proximal colon; not those that are of diminutive size and located in the left colon.26 In our study, only 13 (1.3%) of all the diminutive polyps were sessile serrated adenoma/polyps, and only 6 (0.62%) of which were located in the left colon. None had cytologic dysplasia. Other studies have shown similar low rates of diminutive sessile serrated adenoma/polyp in the rectosigmoid.27 We classified all serrated polyps as non-neoplastic (NICE Type 1), as endoscopic pattern and features of sessile serrated adenoma/polyp are more akin to those of hyperplastic polyps.28 ,29 Current optical diagnosis strategies endorse the resection of all right-sided colon polyps, and thus, there should not be increased apprehension that a right-sided sessile serrated adenoma/polyp would remain in situ if it was erroneously diagnosed as a hyperplastic polyp. Nonetheless, endoscopic and pathologic diagnoses of sessile serrated adenoma/polyps have not shown good reliability. Until such endoscopic and pathologic distinctions are more consistent, it may be necessary to submit all right colon serrated polyps, irrespective of size, to pathology. We attempted to account for this limitation in our study by reporting a low-confidence optical diagnosis if the lesion showed features suggestive of a sessile serrated adenoma/polyp (ie, assignment of low confidence would entail the polyp to be submitted for histopathology analysis).
We performed our study of optical diagnosis comparing near focus view using the latest commercially available endoscope system (Exera III) with standard view using the previous generation endoscope system (Exera II). The magnification (65×) coupled with the increased resolving power (1.56×) in the near focus view compared with the standard view likely had the greatest influence on the displayed image for the optical diagnosis. Other features of the new technology, such as a prefreeze function that identifies the sharpest image from the earlier 4–29 frames may have reduced the number of attempts needed to capture a sharp image for interpretation. The 1080p (progressive scan) in the new technology versus 1080i (interlaced) previously, was less likely to influence our results, as image differences between progressive scan and interlaced systems are notable only with fast-motion video when displayed on a 32-inch, or larger, LCD monitor.
Our positive findings of optical diagnosis of colorectal polyps expand on previous single academic centre studies that showed adequate surveillance interval recommendations.6 ,30 Nonetheless, recent non-academic setting studies failed to achieve high levels of accuracy31 ,32 suggesting that our results may not be immediately generalisable with community practice. It is possible that the interest, knowledge and motivation in learning and practicing the optical diagnosis of colorectal polyps in the non-academic setting may be less due to the lack of both financial and non-financial gains.33 ,34
Importantly, we show that optical diagnosis can be successfully practiced with the proper training,35 classification and tools. Indeed, achievement and sustainment of high performance requires more than one training session.36 A practical self-learning module is available for continued learning.37 Others have described instructor-led training for those without previous knowledge of the use of image enhancement.34 ,38 ,39 Standardisation of training as well as development of a quality assurance auditing system will be important factors in the widespread implementation of the optical diagnosis strategy. Realistically, the greatest challenge may be to motivate the endoscopists to provide an optical diagnosis, perhaps because of the lack of financial incentive or the perceived risk of not sending advanced pathology. Changes in healthcare reimbursement, such as bundled payments, may be a future driver. Digital integration of the polyp image and optical diagnosis into the endoscopy reporting system with photo documentation and archiving would permit both self and formal reviews to reassure providers, patients and payers.
In conclusion, we provide evidence for a paradigm shift in the clinical practice of colonoscopy for colorectal cancer screening from universal pathology to real-time optical diagnosis of diminutive colorectal polyps. Use of colonoscopes capable of near focus view can lead to an increased number of high-confidence optical diagnoses of diminutive colorectal polyps with high accuracy and NPVs. Using optical diagnosis, patients may have diminutive polyps diagnosed and removed at the time of the procedure, and the surveillance intervals immediately relayed, with significant potential in cost savings.
Acknowledgments
The authors would like to thank Aparna Motiwala and Sydney Johnson for their diligent efforts and work in patient enrolment, data collection and management; Alex McMillan, PhD, Biostatistics, Health Research and Policy, Stanford University, for his invaluable statistical assistance in the study design, randomisation and analysis; and the GI endoscopy units nursing and technical staff. The study was partially supported by Olympus Medical America. The contents of this article do not represent the views of the Department of Veterans Affairs or the United States Government.
References
Footnotes
Contributors TK: study concept and design; acquisition, analysis, and interpretation of data; drafting the work and critical revision of the manuscript for important intellectual content; final approval of the version to be published; agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. AR, RVR and KM: study concept and design; acquisition and interpretation of data; critical revision of the manuscript for important intellectual content; final approval of the version to be published; agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. TS: acquisition and interpretation of data; critical revision of the manuscript for important intellectual content; final approval of the version to be published; agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. AB: acquisition and interpretation of data; critical revision of the manuscript for important intellectual content; final approval of the version to be published; agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. JCK: interpretation of data; critical revision of the manuscript for important intellectual content; final approval of the version to be published; agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. RS: study concept and design; acquisition and interpretation of data; drafting the work and critical revision of the manuscript for important intellectual content; final approval of the version to be published; agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Funding Olympus Medical America, Inc. provided partial funding. Neither investigators nor patients received financial compensation for the trial. The investigators designed the trial, gathered, analyzed and maintained the data, and decided to submit the manuscript for publication. The first and senior authors drafted the manuscript, and all authors contributed to its revision. The sponsor had no role.
Competing interests These authors disclose the following: TK, AR and RS have received research funding from Olympus Medical America and are consultants for Olympus Medical Systems Corporation. No other conflicts of interest to disclose. Study participants were advised of potential investigator conflicts of interest.
Ethics approval Stanford University IRB & Veterans Affairs Palo Alto Research.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement All authors had access to the study data. The sponsor had no influence on data management, analysis and publication.