5 Committee discussion
5.1 The committee discussed current practice for assessing suspected cervical abnormalities in a colposcopy clinic. The clinical experts explained that NHS clinics most often use binocular colposcopy, which allows a colposcopist to examine a cervix and take both diagnostic and treatment biopsies under direct visualisation. Acetic acid is used to highlight areas of abnormality. The committee noted that colposcopy is associated with both intra- and inter-observer variability because it is a visual examination that is highly dependent on the colposcopist's expertise. The committee considered the role of the adjunctive colposcopy technologies and was advised by the clinical experts that the technologies could provide less subjective results and help colposcopists select areas for biopsy. The clinical experts also explained that the technologies could help identify high-grade lesions in people referred for colposcopy because of low-grade cytology.
5.2 The committee noted that a series of changes are being made to the screening pathways used in the NHS cervical screening programme (NHSCSP). Human papilloma virus (HPV) triage was fully implemented in England in April 2014. HPV primary screening is currently being done in several pilot sites, with full implementation in England expected in 2019. These changes could affect referrals to colposcopy clinics and consequently the prevalence of high-grade disease, particularly when people with a HPV-positive cytology-negative screening result are seen in colposcopy. The committee concluded that there had been substantial changes to the care pathways since NICE's first diagnostics assessment of the DYSIS colposcope with DYSISmap in 2012.
5.3 The committee discussed the external assessment group's (EAG) critical appraisal of the included diagnostic accuracy studies. It noted that the greatest risk of bias in the studies occurred because not all patients had the reference standard test (colposcopically directed biopsies and histopathology). In most studies, people who had a negative colposcopy did not have biopsies taken. The clinical experts explained that it was not considered good clinical practice to take biopsies when there was no clinical indication. But the committee noted that the EAG's sensitivity analyses on the effect of verification bias showed that the more random biopsies taken, the lower the estimates of both sensitivity and specificity. The committee concluded that the diagnostic accuracy estimates provided by the included studies were likely to have been influenced by verification bias, and highlighted that future studies should aim to minimise this when possible.
5.4 The committee considered the applicability of the diagnostic accuracy studies that were done outside the UK. The clinical experts explained that the quality assurance measures for colposcopy done outside the UK are different to those in the UK, and that this was likely to influence the accuracy of colposcopy. The committee noted that the NHSCSP recommends that a satisfactory colposcopy should have a 65% positive predictive value for CIN 2+. It considered that although positive predictive value was likely to be influenced by several confounding factors, video colposcopy in the DYSIS studies did not achieve this benchmark, with a pooled positive predictive value of 55.78%. However, the committee noted that because this value depends on disease prevalence, the use of positive predictive value to assess the generalisability of studies to UK practice is problematic. The clinical experts also noted that the pooled sensitivity of colposcopy in the DYSIS studies was lower than they would expect to see in the UK. They also noted that in the ZedScan I study, which was done in the UK and used binocular colposcopy, a higher sensitivity for colposcopy was reported. The committee concluded that because of differences in colposcopy practice, such as fewer quality assurance measures and the use of video colposcopy, the accuracy data from non-UK studies may not be generalisable to the NHSCSP.
5.5 The committee considered the potential for the adjunctive colposcopy technologies to reduce both intra- and inter-observer variability. The companies explained that the technologies are designed to reduce the subjectivity of colposcopy by providing more objective results, but noted that no data on the reproducibility of the tests had been presented for the assessment. However, the committee noted published data suggesting that clinicians felt that the DYSISmap improved their confidence when selecting biopsy sites. It concluded that the technologies had the potential to help standardise colposcopy examinations, but that insufficient data were available to determine whether this benefit would be realised in NHS clinical practice.
5.6 The committee discussed the results of the diagnostic accuracy analyses for the DYSIS colposcope with DYSISmap and the ZedScan I. It noted that although the accuracy estimates for colposcopy alone in the DYSIS and ZedScan studies varied considerably, the estimates suggested that the technologies were more sensitive but less specific than colposcopy alone. It considered that in practice this would result in a reduced false negative rate with more people being diagnosed with CIN 2 or worse (CIN 2+). But this could be at the expense of a higher false positive rate with more people having unnecessary diagnostic biopsies and treatment. The committee further noted that the diagnostic odds ratios, which had been calculated by the EAG for the DYSIS colposcope with DYSISmap studies, suggested that there was no difference between the accuracy of DYSIS colposcopy alone and DYSIS colposcopy with DYSISmap. The committee concluded that the results of the diagnostic accuracy studies suggest that it is plausible that the adjunctive colposcopy technologies may change the test threshold so that more people have biopsies, but without improving colposcopists' ability to differentiate between high- and low-grade disease.
5.7 The committee discussed the Cholkeri-Singh et al. (2018) and DeNardis et al. (2017) studies, submitted as prepublication manuscripts during consultation. These provided data from the IMPROVE-COLPO study. It acknowledged that these studies provide real world outcome data on the number of biopsies taken and supplement the diagnostic accuracy data in the EAG's systematic review. The committee noted that the results of the Cholkeri-Singh et al. study show that DYSIS with DYSISmap detects additional cases of both CIN 2 and CIN 3, relative to standard colposcopy, without increasing the number of people having biopsies. The committee considered the design of this study and noted a lack of detail on the methods used to ensure that controls in the retrospective arm were comparable with the people in the prospective arm. However, the committee heard from the EAG that the people in the 2 study arms appear to be comparable for key baseline characteristics. The committee also considered analyses provided at consultation based on KC65 data (from the NHSCSP in England between 2012/13 and 2015/16). It noted that the data generally showed no increase in biopsy rates in centres adopting DYSIS, but it acknowledged that DYSIS may not be used for every colposcopy in these centres. The committee concluded that despite these papers having methodological limitations, combined with the KC65 data they provided some reassurance that the increase in biopsies implied by the results of the diagnostic accuracy studies alone may not be realised in practice in centres using DYSIS colposcopy with DYSISmap.
5.8 A patient expert explained that referral for colposcopy can often cause substantial anxiety, which may not reduce even when the colposcopy is normal. People having a colposcopy may be anxious because of the examination itself and because they have already had a screening result informing them that an abnormality has been detected. The clinical experts explained that it can often be difficult to reduce anxiety in people who have a negative colposcopy, but who were referred with an HPV-positive result, because no treatment can be offered. The committee noted evidence from the systematic review and also anecdotal evidence from clinical and patient experts, which suggested that the adjunctive colposcopy technologies could reduce anxiety because people can be shown objective information to explain that no abnormality has been detected. The committee concluded that although the additional information provided by the adjunctive colposcopy technologies has the potential to help clinicians reassure people and reduce their anxiety, there are currently insufficient data to conclude that they have a significant effect on this (see section 6.3).
5.9 The committee discussed the assumption made in the cost-effectiveness model that video colposcopy and binocular colposcopy are equivalent in terms of diagnostic accuracy. The clinical experts explained that there was no consensus among experts about their equivalence and that the sensitivity estimates for video colposcopy obtained in the DYSIS studies were lower than would be expected for binocular colposcopy in the NHS. Also, the clinical experts noted that the estimates for the sensitivity of binocular colposcopy in the ZedScan studies were higher, and more representative of NHS practice. But the committee noted that the estimates used in the cost-effectiveness model for colposcopy alone were taken from the meta-analyses of DYSIS colposcopy. Therefore, the committee concluded that the relative benefits of the adjunctive colposcopy technologies could have been overestimated in the modelling.
5.10 The committee discussed both modelled base cases and noted that the increased sensitivity of the adjunctive colposcopy technologies led to less cervical cancers developing over the 60‑year time horizon. The clinical experts explained that the additional high-grade lesions detected using the adjunctive colposcopy technologies could in fact be low-volume CIN 2 disease, which could regress without treatment. The committee questioned whether data were available that explained the natural history of low-volume CIN 2 but heard that these were not available. Anecdotal evidence, and results of a British Society for Colposcopy and Cervical Pathology survey, suggest that some clinicians are now using either ablative techniques or 'watchful waiting' management strategies for low-volume CIN 2 in some circumstances. The committee also noted that when the time horizon of the model was reduced to 3 years, and the longer-term outcomes associated with increased sensitivity were removed, colposcopy alone dominated; that is, it was more effective and less expensive than the adjunctive colposcopy technologies. The committee concluded that, without clinical outcome data, or data on the natural history of low-volume CIN 2, there was uncertainty about the longer-term outcomes associated with the increased sensitivity of the adjunctive colposcopy technologies. It wished to encourage further data collection to resolve this (see section 6.4).
5.11 The committee discussed the effect of the lower specificity associated with the adjunctive colposcopy technologies on longer-term outcomes in the model. In the shorter term, the model showed that reduced specificity is associated with an increase in unnecessary biopsies and treatments. The committee questioned whether this would be realised in practice. The EAG advised that the assumptions made in the model about when biopsies would be taken were based on the NHSCSP's colposcopy and programme management guidelines (2016; publication number 20). The clinical experts explained that these guidelines may not always be followed, and colposcopists may take biopsies for reassurance that high-grade disease is not present. The committee noted that there is considerable variation in clinical practice between colposcopists, and that there were no data to show how the adjunctive colposcopy technologies affect clinical decision-making in the UK. The committee also noted its previous conclusion (see section 5.7) that results from the prepublication version of Cholkeri-Singh et al. (2018) study and the KC65 data (from the NHSCSP in England between 2012/13 and 2015/16) showed no increase in biopsy rates in centres adopting DYSIS. It also noted that Cholkeri-Singh et al. reported that using DYSIS was associated with an increased yield of CIN 2+ which, combined with the data on biopsy rates, suggests that DYSIS helps colposcopists target the areas chosen for biopsy. The committee concluded that there is some real world evidence suggesting that DYSIS does not increase the biopsy rate to the extent predicted by the model, and noted that equivalent data were not yet available for ZedScan I.
5.12 The committee discussed whether reduced specificity is associated with an increased risk of adverse obstetric outcomes in the longer term. The clinical experts explained that the relationship between biopsies, treatment and adverse obstetric outcomes was not well understood, but it was generally acknowledged that the smaller the excisional treatment the lower the risk of adverse outcomes. The committee noted that the base case assumed an excess risk of preterm delivery of 0.04, which was reduced to 0 in a scenario analysis with no substantial effect on the results. The committee concluded that although they were an important clinical consideration in practice, the longer-term effects of reduced specificity did not seem to be a key driver in the model.
5.13 The committee questioned the cost savings of the adjunctive colposcopy technologies in the model. The EAG explained that the model's cost savings were driven by increased sensitivity, which led to a reduction in costs associated with both cancer treatment and follow-up appointments. The clinical experts noted that technologies that improve the negative predictive value of colposcopy may become more important after HPV primary screening is fully rolled out and people with HPV-positive, cytology-negative results are referred for colposcopy. The committee noted that the base case for HPV primary screening was based on preliminary data only, but acknowledged that improvements in sensitivity may become increasingly important in the future. The committee concluded that because there were no data on the natural history of low-volume disease, it was uncertain whether the adjunctive colposcopy technologies would increase the detection of disease that would progress to cancer if not treated. Therefore, the cost savings in the model may not be robust.
5.14 The committee questioned the effect of not having a probabilistic sensitivity analysis to quantify the overall uncertainty in the model. The EAG explained that it could not do this analysis because of the length of time needed to run each simulation. The EAG also explained that although the mean ICER may be different from the deterministic analyses if the model was run probabilistically, there was unlikely to be a substantial difference that would change the modelling conclusions. The committee noted that the model results had been robust to changes in many parameter estimates and assumptions in the deterministic sensitivity and scenario analyses, but that the results were likely to be confounded by the lack of clinical outcome data. The committee concluded that on this occasion the lack of a probabilistic sensitivity analysis was not critical.
5.15 The committee considered whether the adjunctive colposcopy technologies should be recommended for routine adoption. It noted its conclusions on the applicability of data from non-UK studies where the accuracy of colposcopy may differ (see section 5.4), the lack of data on the natural history of low-volume CIN 2 (see section 5.10) and the uncertainty about whether the adjunctive colposcopy technologies would reduce cervical cancer over the longer term (see section 5.13). Taking these factors into account, the committee considered that there was uncertainty about the clinical and cost effectiveness of the adjunctive colposcopy technologies because only diagnostic accuracy data were available. It noted, however, that further data (prepublication versions of Cholkeri-Singh et al. 2018 and DeNardis et al. 2017) provided at consultation showed that DYSIS was able to detect more CIN 3 lesions than standard colposcopy, without increasing the number of people having biopsies. Therefore, the committee concluded that there was enough evidence that colposcopy using DYSIS with DYSISmap detects more clinically important lesions than colposcopy alone to recommend its continued adoption. It also noted that the additional data provided at consultation were from a US study. The committee wished to encourage centres using DYSIS to audit their outcomes and confirm that the expected benefits are achieved in the NHS (see section 5.16). Also, the committee concluded that although the ZedScan I shows promise, there was too much uncertainty over clinical and cost effectiveness to recommend its routine adoption at present, and recommended that further research was needed (see section 5.17).
5.16 The committee recalled that the available clinical outcome data that support using DYSIS with DYSISmap were from a US study (see section 5.15). It therefore recommended that centres using this technology should audit their clinical outcomes and confirm that the expected benefits are achieved in the NHS. Outcomes that should be audited include, but are not limited to, rates of CIN 2+ detection, CIN 3+ detection and biopsy.
5.17 The committee considered the amount of evidence available for both adjunctive technologies. It noted that more data were available for the DYSIS system, and that evidence for ZedScan I was limited to a small number of diagnostic accuracy studies. The committee considered that ZedScan I shows promise but further studies are needed, in particular to compare the accuracy and the clinical effectiveness of the technology with standard colposcopy.
5.18 The clinical experts explained that all colposcopy clinics complete a quarterly data return for Public Health England, the KC65. This is used to compare and assess their data against the standards outlined in NHSCSP's colposcopy and programme management guidelines (2016). The committee considered whether these data could be studied to see if biopsy and detection rates of CIN 2+ had increased in centres that had already adopted DYSIS colposcopy with DYSISmap or the ZedScan I. The clinical experts explained that the device used in each colposcopy is not currently recorded and it is not known whether centres with an adjunctive colposcopy technology use it routinely. The committee acknowledged that making the necessary changes to the KC65 to collect these data would be difficult. However, it wished to encourage the owners of the KC65 dataset to consider whether it could be adapted in the future and used to support further data collection for the adjunctive colposcopy technologies, and whether papers based on the data could be published and used for updates of this guidance. The committee also suggested that, if it is not possible to use the KC65 to collect these data nationally, then local audits should be used to collect these data from services that have adopted the adjunctive technologies.
5.19 The committee identified that different thresholds had been used to assess the accuracy of colposcopy in the studies. Some used colposcopic impression that high-grade disease was present (that is, what the colposcopist thought). In other studies colposcopy was considered positive if at least 1 measurement point was suggested for biopsy (that is, what action was taken). Some studies used both. This made comparison of the relative cost effectiveness of the adjunctive technologies difficult. The committee considered that a more consistent approach to assessing and reporting colposcopic accuracy in studies would help future comparisons of adjunctive technologies. The clinical experts stated that work on producing standards for reporting colposcopic accuracy in the NHSCSP is being done.
5.20 The committee noted the assumption made in the cost-effectiveness model that video colposcopy and binocular colposcopy are equivalent in terms of diagnostic accuracy (see section 5.9). The clinical experts explained that there is limited evidence to support this assumption. Future assessments of adjunctive technologies would benefit from research assessing the equivalence of different types of colposcope (digital, video and binocular).