Virtual chromoendoscopy to assess colorectal polyps during colonoscopy

The committee considered the generalisability of the evidence base to clinical practice in the NHS. The committee noted that most of the endoscopies in the studies included in the assessment were done by experienced endoscopists in single academic centres, most of which were outside of the UK. The committee also noted that the UK‑based DISCARD 2 study was excluded from the assessment because only 22% of the participating centres had high-definition equipment. The committee heard from clinical experts that DISCARD 2 was a multicentre community-based study, with 28 endoscopists, which compared Narrow Band Imaging (NBI) with histopathology and was considered to reflect clinical practice in the NHS. The results of this study showed that the sensitivity of NBI for real-time assessment of diminutive polyps was lower than the accuracy estimated in this assessment (0.76 compared with 0.87 to 0.92). The committee concluded that the diagnostic accuracy of virtual chromoendoscopy technologies reported in this assessment reflect the accuracy that could be achieved by endoscopists with experience of using virtual chromoendoscopy and who work in specialist or academic settings. The committee concluded further that diagnostic accuracy results probably do not reflect the accuracy that would be achieved by endoscopists with limited experience of virtual chromoendoscopy and who work in community-based settings.

The committee considered the differences between the 3 virtual chromoendoscopy technologies (NBI, flexible spectral imaging colour enhancement [FICE] and i‑scan). The committee heard from clinical experts that FICE and i‑scan work differently to NBI; they are software-based image enhancement technologies, whereas NBI uses optical filters on white light, resulting in narrow-band light which enhances the contrast between the vessels and the surrounding mucosa. The committee also heard that the type of technology in place in centres is likely to vary, and equipment is replaced every 5 to 8 years. The committee then considered the different levels of evidence available for NBI, FICE and i‑scan. It noted that most studies were on NBI and very few studies were on FICE and i‑scan. It also noted that most of the studies on i‑scan were done in academic centres, by 1 endoscopist experienced in using virtual chromoendoscopy, and this resulted in higher accuracy results for i‑scan compared with NBI. It noted also that none of the studies on FICE limited the accuracy data to high-confidence characterisations of polyps, and this resulted in lower accuracy results for FICE compared with NBI. The committee concluded that, without direct comparative data, it is unclear whether one virtual chromoendoscopy technology is superior to others. It concluded further that NBI, FICE and i‑scan will probably perform similarly in clinical practice, because the diagnostic accuracy achieved is likely to depend on the experience level of the endoscopist and the level of confidence in the polyp characterisation more than on the virtual chromoendoscopy technology used.

The committee considered the diagnostic accuracy of virtual chromoendoscopy technologies for real-time assessment of diminutive polyps. The committee noted that the American Society for Gastrointestinal Endoscopy has developed criteria on diagnostic accuracy that endoscopic technologies must meet before being considered appropriate for use in US clinical practice (the Preservation and incorporation of valuable endoscopic innovations [PIVI] criteria). The PIVI criteria on real-time assessment of diminutive colorectal polyps guides decisions on resecting and discarding polyps without histopathologic assessment. These criteria are:

• technologies should have an agreement of 90% or more with the surveillance intervals set by histopathology

• the negative predictive value of the technology for assessing adenomatous polyp histology should be 90% or more.
  
  The committee heard from clinical experts that the PIVI criteria, which are used in US clinical practice, were widely accepted in the UK gastrointestinal community. The committee concluded that the diagnostic accuracy of NBI, FICE and i‑scan were likely to meet the PIVI criteria if used by endoscopists with experience of virtual chromoendoscopy technologies.

The committee discussed the accuracy of the comparator test, histopathology. The committee heard from clinical experts that histopathology is considered to be the gold standard in current practice, but it is actually an imperfect reference standard for diagnosing polyps. The committee also heard from clinical experts that currently about 8% to 10% of diminutive polyps do not have histopathology assessment because they are lost or destroyed before they reach the histopathologist and they are therefore assumed to be adenomatous. It heard further that polyp characterisation using histopathology assessment is 90% to 95% correct. The committee concluded that given the limitations of histopathological assessment of polyps, the diagnostic accuracy of the virtual chromoendoscopy technologies is likely to be more accurate than data from the studies suggests.

The committee discussed the consequences of misdiagnosing diminutive polyps using virtual chromoendoscopy. The committee noted that if virtual chromoendoscopy is used for real-time assessment of polyps, 3% to 6% of the surveillance intervals are likely to be incorrectly assigned. The committee heard from clinical experts that if virtual chromoendoscopy is used, over-surveillance would be slightly more common than under-surveillance. The committee noted that the effect on clinical outcomes from incorrectly leaving diminutive adenomatous polyps in place and incorrectly assigning a surveillance interval that is too long is uncertain. The committee heard from the external assessment group (EAG), however, that the lifetime risk of colorectal cancer calculated from the model was similar for the 3 virtual chromoendoscopy technologies and histopathology (3.025% for histopathology, 3.020% for NBI, 3.045% for FICE and 3.021% for i‑scan; see section 4.44). It concluded that although there was some uncertainty over how the diagnostic accuracy data would translate into clinical outcomes, it was aware that an end-to-end study on clinical outcomes would need to be done on a large cohort over a long period of time and so may not be feasible.

The committee discussed the uncertainties around using the School of Health and Related Research's (ScHARR) bowel cancer screening (SBCS) model for the assessment. The committee was aware that ScHARR ran the SBCS model on behalf of the EAG, and therefore the EAG was unable to internally validate the model results. However, it noted that the model had previously been validated for use to inform the NHS bowel cancer screening programme strategy, and that the costs in the model had been updated to reflect current costs. The committee heard from the EAG that there were structural uncertainties in the model, for example, the accuracy of virtual chromoendoscopy was not used for ongoing surveillance. However, the committee noted that it would not have been possible for the EAG to build a de novo model because of the level of resource needed to develop such a complex model. The committee therefore concluded that although there was some uncertainty about the SBCS model's results, it was considered to be the most appropriate model for the assessment.

The committee considered the cost of histopathology assessment of polyps used in the model. It heard from the EAG that in the base-case analysis, the cost of histopathology per polyp was based on the NHS reference cost for direct access pathology for 2014/15, which lists the cost of histopathology and histology as £28.82 (DAPS02). The committee noted that this reference cost is likely to include requests from community services, such as GPs, for histopathology and that there is no stratification by sample type (for example, type of specimen or tissue preparation), which may affect the cost. The committee noted further that the true cost of histopathology assessment of colorectal polyps was probably more than £50 per polyp. The committee concluded that the cost of histopathology was likely to be underestimated in the model, and so the cost savings for virtual chromoendoscopy technologies were likely to be greater than the model suggested.

The committee discussed the proportion of hospitals that already have high-definition enabled virtual chromoendoscopy equipment in place. The committee heard from the EAG that the economic model assumed that the cost of upgrading colonoscopy equipment would be included in the NHS reference costs for colonoscopy (see table 3). The committee heard from clinical experts that most endoscopes were replaced every 5 to 8 years and the video system is likely to be replaced every 10 years because repairs after this period are often not supported. The committee heard further that most centres will have at least 1 virtual-chromoendoscopy-enabled machine. The committee concluded that the assumption made in the economic model was reasonable.

The committee discussed the assumption used in the model that histopathology is 100% accurate when assigning surveillance intervals. It heard from clinical experts that although histopathology is considered to be the gold standard, the diagnostic accuracy is likely to be below 100% (see section 5.5). The committee concluded that the clinical effectiveness of histopathology was likely to have been overestimated in the model, and therefore the difference in clinical effectiveness between histopathology and the virtual chromoendoscopy technologies was likely to be smaller than the results suggested.

The committee considered the implications for histopathology laboratories of adopting virtual chromoendoscopy for real-time assessment of colorectal polyps. The committee heard from clinical experts that histopathology laboratories are under considerable strain because of high workloads, and that diminutive colorectal polyp assessment is an important cause of this overload. The committee discussed whether using virtual chromoendoscopy for real-time assessment of diminutive polyps rather than sending all of these to histopathology could reduce this workload and result in cost savings or free histopathologists for other priorities. The committee noted that the endoscopist's level of experience would affect how many diminutive polyps are assessed with high confidence, and therefore how many polyps are sent to histopathology. For example, risk-averse practice (in which polyps that are likely to be hyperplastic are removed and sent to histopathology) is probably more common in endoscopists with less experience. Therefore, cost savings through avoiding histopathology assessment may not be as large in this group compared with experienced endoscopists, who are likely to assess more polyps with high confidence and send fewer to histopathology. The committee concluded that virtual chromoendoscopy used by experienced endoscopists could reduce the number of diminutive polyps sent to histopathology laboratories, therefore freeing histopathology resources.

The committee discussed the results of the cost-effectiveness analysis and noted that in the base case, the NBI and i‑scan dominated histopathology, that is, they were cheaper and more clinically effective than histopathology. The committee also noted that in the base case, FICE could be considered cost effective with an incremental cost-effectiveness ratio (ICER) of £671,000 saved per quality-adjusted life year (QALY) lost (see section 4.42). However, the committee noted that the base-case analysis only included adverse events for colonoscopy with polypectomy. The committee heard from a clinical expert that there is also a risk of adverse events from a colonoscopy even without a polypectomy. It heard from the EAG that an analysis was done which included the risks of adverse events from all colonoscopies as well as for colonoscopy with polypectomy (see section 4.51). The committee noted that in this analysis, NBI and i‑scan still dominated histopathology and the ICER for FICE decreased to £342,000 saved per QALY lost. The committee concluded that the most plausible results came from the scenario analysis that included a risk for adverse events for colonoscopy without polypectomy. The committee further concluded that NBI, FICE and i‑scan could be cost-effective options for assessing diminutive polyps.

The committee discussed the robustness of the results of the economic model. It noted that results of the sensitivity and scenario analyses showed that NBI and i‑scan were dominant compared with histopathology in all scenario analyses. It noted further that FICE dominated histopathology in some analyses and was considered cost effective in other analyses, with ICERs ranging from £126,000 to £1,270,000 saved per QALY lost. The committee considered that although there were limitations and uncertainties in the economic assessment (see section 5.7), the sensitivity analyses showed that the results were robust to changes. The committee concluded that the results of the economic model could be considered to be fairly robust.

The committee considered all its discussions on virtual chromoendoscopy, and noted its conclusions that:

• optical diagnosis using virtual chromoendoscopy technologies was likely to meet the PIVI criteria if used by endoscopists with experience of virtual chromoendoscopy technologies (see section 5.4)

• the lifetime risk of colorectal cancer was estimated to be similar when diminutive polyps were assessed and surveillance intervals were set using virtual chromoendoscopy technologies or histopathology (see section 5.6)

• assessment of diminutive colorectal polyps with virtual chromoendoscopy technologies is cost effective compared with assessment of diminutive colorectal polyps using histopathology (see sections 5.12 and 5.13)

• the virtual chromoendoscopy technologies are cost saving when they are used to implement a management strategy which reduces the number of diminutive polyps sent for histopathological analysis (see section 5.11).
  
  The committee therefore concluded that virtual chromoendoscopy using NBI, FICE or i‑scan to assess diminutive polyps during colonoscopy, instead of sending polyps to histopathology, could be considered clinically effective and cost effective if done by a specialist group, that is, endoscopists with expertise in optical diagnosis using virtual chromoendoscopy technologies.

The committee considered whether using virtual chromoendoscopy for real-time assessment of diminutive polyps and using a discard strategy was acceptable to people. The committee heard from a clinical expert that there were no UK‑based studies that looked at patient acceptability, but 2 studies from the US and 1 study from Australia with data on patient acceptability were available. In the US study, many patients stated that they would pay $150 from their own pocket to have polyps removed and assessed by histopathology, instead of using real-time assessment of polyps with a discard strategy (Vu et al. 2015). The committee concluded that further research on patient acceptability of virtual chromoendoscopy for real-time assessment of diminutive polyps and use of a discard strategy would be valuable.

The committee considered the effect of training for endoscopists on the diagnostic accuracy of NBI, FICE and i‑scan. The committee heard from clinical experts that the DISCARD 2 study had implemented a programme consisting of a 1‑hour training session using PowerPoint images followed by a test. The committee noted that the results of the study suggested that training and monitoring for endoscopists needed to be more rigorous to maintain high levels of diagnostic accuracy for virtual chromoendoscopy technologies. The committee heard that the manufacturers of the technologies offer 2 forms of training for endoscopists, both developed with experts: peer-to-peer training at centres of excellence; and online training for self-study. The committee also heard that general experience in diagnosing polyps and familiarity with polyp classification systems, combined with acting on feedback from peers, were important factors in improving the skill levels of endoscopists. It concluded that the most effective forms of training should be determined, and that this could be done through collaboration between manufacturers of virtual chromoendoscopy technologies and professional organisations.

The committee discussed the need for quality assurance measures to be in place before virtual chromoendoscopy for assessment of polyps during colonoscopy can be used in clinical practice. It heard from clinical experts that the skills of endoscopists who do colonoscopies are known to vary. The committee heard further that quality assurance measures, such as accreditation and monitoring of practice, were needed to ensure that virtual chromoendoscopy for making optical diagnoses is only used by endoscopists who can meet the PIVI criteria, and to maintain high levels of diagnostic accuracy over time. The committee also noted that there was currently no accreditation or monitoring system in place for virtual chromoendoscopy and heard that any accreditation and monitoring scheme would need to be rolled out to both clinicians and nurse-endoscopists. The committee concluded that a national accreditation scheme for using virtual chromoendoscopy to make optical diagnoses should be developed. It concluded further that when virtual chromoendoscopy technologies are used, intermediate measures should be monitored for quality assurance and to give endoscopists ongoing feedback.