VivaScope® 1500 and 3000 systems for detecting and monitoring skin lesions: a systematic review and economic evaluation

Interventions

• Diagnosis Assessment of the lesion by dermoscopy plus VivaScope or VivaScope alone by an experienced skin cancer specialist.

• Delineation of lesion margins Assessment of the lesion by dermoscopy plus VivaScope or VivaScope alone by an experienced skin cancer specialist.

Although this report is mainly aimed at the current versions of VivaScope (1500 and 3000), earlier versions such as VivaScope 1000 and 2500 were also considered, as they may provide additional potential information on the current versions.

Comparators

• The comparator eligible for inclusion for the assessment of both diagnostic accuracy and delineation of lesion margins was visual assessment of the lesion followed by dermoscopy and clinical judgement by an experienced skin cancer specialist.

Reference standard

• The eligible reference standard for the assessment of diagnostic accuracy and margin delineation was histopathology or biopsy of the excised skin lesion.

Study design

The following types of studies were eligible for inclusion:

• randomised controlled trials (RCTs) or observational studies, in which participants are assigned to dermoscopy plus VivaScope or VivaScope alone for diagnosis or skin lesion delineation, and where outcomes are compared at follow-up

• test accuracy studies assessing the test accuracy of dermoscopy plus VivaScope or VivaScope alone with histology of biopsy as the reference standard.

The following study/publication types were excluded:

• preclinical and animal studies

• reviews, editorials and opinion pieces

• case reports.

Included studies

A total of 7446 records were identified from clinical effectiveness searches in electronic databases. After deduplication, 5122 records were screened for eligibility based on title and abstract (Figure 1).

FIGURE 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram for studies included and excluded from the clinical effectiveness review.

Full publications of 347 references were ordered and, after screening for eligibility, 11 studies30–40 met the inclusion criteria. The database searches were updated from October 2014 to February 2015, and a further two studies41,42 that met the inclusion criteria were identified. Three additional studies43–45 were obtained by hand-searching and contacting clinical experts in the field. Thus, in total, 1630–41,43–46 studies were identified that met the inclusion criteria for the review. No study was identified from conference proceedings that met the inclusion criteria.

Figure 1 shows the flow diagram for included and excluded studies of clinical effectiveness. A list of excluded references (with reason for exclusion) is presented in Appendix 2, and Appendix 3 shows a list of ongoing trials identified from searching trial registers.

Study characteristics

Quality assessment of studies included in clinical effectiveness review

The QUADAS-2, which separates the evaluation of study quality into two main areas – (1) risk of bias and (2) concerns regarding applicability of patient selection, index test, reference standard and flow of timing – was used to assess quality of included studies.

A summary of the results of the quality assessment of the included studies is shown in Appendix 4. The majority of the included studies had a low risk of bias and low applicability concerns in patient selection (e.g. less concern that included patients did not match the review question),30,32–34,36,38,40–45 conduct of the index test (e.g. the index test, its conduct or interpretation did not differ from the review question)30–34,36–38,40,41,43–45 and reference standard. 31–36,38,40–45 However, concerning flow and timing, the risk of bias in the majority of the studies was unclear (i.e. it was unclear if patient flow did not introduce any bias or also if there was an appropriate interval between the index test and reference standard)32,33,35–45 as a result of poor reporting and/or insufficient data.

Figure 2 shows a summary of risk of bias and applicability concerns of included studies.

FIGURE 2.

Summary of risk of bias and applicability concerns of included studies.

Lesion diagnosis

Lesion margin delineation

Lesion diagnosis

None of the included studies on lesion diagnosis reported lesion recurrence data.

Lesion margin delineation

In the trial conducted by Guitera et al. ,35 none of the patients treated surgically after histopathology confirmed LM (n = 17) or LMM (n = 37) developed recurrence during a median follow-up of 37 months. Recurrence was suspected in one imiquimod-treated patient after 1 year’s follow-up, and in three patients treated with radiotherapy (one after 12, one after 24 and one after 36 months’ follow-up) (Table 17).

Lesion diagnosis

Lesion margin delineation

The only included study on lesion margin delineation2 did not report on misdiagnosis or misclassification of lesions.

Lesion diagnosis

No included study on lesion diagnosis reported change in management of lesions after diagnosis.

Lesion margin delineation

In the trial conducted by Guitera et al. ,35 VivaScope 1500 mapping changed the management of lesions in 27 patients (73%): 11 patients had a major change in their surgical procedure and 16 were offered radiotherapy or imiquimod treatment. Treatment was surgical in 17 out of 37 patients.

Economic evaluations

The systematic literature search identified a total of 125 papers. Of those, 91 were excluded on the basis of title and abstract and 29 were duplicates. Therefore, a total of five papers were identified as potentially relevant and were ordered for full review based on the criteria listed in Box 2. Of the five papers ordered, none was considered to meet the predefined inclusion criteria listed in Box 2. Reasons for exclusion of the ordered papers are provided in Appendix 5.

During the development of this report, the company made available to the EAG an unpublished study of the cost-effectiveness of RCM in the diagnosis of skin lesions suspicious for skin cancer (note that this report has now been published). 46 The study had a retrospective design and, therefore, did not meet the inclusion criteria for economic evaluations described in Box 2. Nevertheless, because of the paucity of any relevant economic evidence on the cost-effectiveness of VivaScope, it was decided to relax the inclusion criteria and thus include this study in the systematic literature review. None of the five potentially relevant papers that had been excluded according to the predefined inclusion criteria met the relaxed inclusion criteria. Figure 3 provides the flow chart of the process of the systematic search for economic evaluations.

FIGURE 3.

Flow chart of the process of the systematic search for economic evaluations. HTA, Health Technology Assessment; NHS EED, NHS Economic Evaluation Database.

The study by Pellacani et al. 46 assessed the cost-effectiveness of RCM for the diagnosis of melanomas in Italy, under a hospital perspective, by estimating the impact of RCM use on the number of benign lesions needed to excise a malignant melanoma, in terms of clinical outcomes and costs per patient. The study, which had a retrospective design, utilised data from the whole skin cancer activity performed in the Province of Modena in a semester, between 1 January and 30 June 2013, and compared costs and outcomes (in terms of the number of skin neoplasms that were excised) between a territorial service that used exclusively dermoscopy for the diagnosis of malignant skin lesions and a university hospital that used a combination of dermoscopy and RCM for the evaluation of suspicious skin lesions.

Clinical data on number of cases, diagnoses and excision procedures were retrieved from two sources: the digital archive of the province-centralised pathology department, which contained data on all the skin tumours that were excised during the study time frame at the university hospital and the territorial dermatology service; and the database of the melanoma and pigmented lesion clinic of the university hospital. The data of the two databases were matched for consistency.

The cost analysis was performed following a microcosting approach. All procedure costs were estimated by taking into account required staff time and other fixed and variable costs (including device, disposables, etc.). The cost of RCM examination included depreciation and other fixed and variable costs, assuming an operational lifetime of 4 years and the use of the RCM on at least 1800 patients per year. Overhead costs were evaluated for all procedures at 15% of the unit total direct costs. Unit costs (staff, consumables) were taken from the University Hospital of Modena. Training costs for clinicians assessing RCM images were not included. For lesions not excised after RCM examination, the cost of a dermoscopy follow-up examination was added. A surgical excision because of dermoscopy changes was estimated for 14% of lesions referred to digital follow-up. Other direct and indirect costs not directly relevant to the diagnostic process, such as time off work for accessing the hospital services, transportation, morbidity, etc., were not considered in the cost analysis.

The study estimated a large reduction in the number of benign lesions excised at the university hospital (which used RCM in addition to dermoscopy) compared with the territorial dermatology service (which used exclusively dermoscopy), with almost the same number of melanomas excised in both services. The analysis of data showed a number needed to excise (NNE) of 6.25 for the university hospital, compared with 19.41 for the territorial dermatology service.

In terms of costs, the study estimated a mean total cost per patient following the standard diagnostic procedure without RCM of €144, including examination with dermoscopy (70%) or digital dermoscopy (30%), excision, medication, pathology report and conclusive visit. Applied to all cases undergoing excision in the scenario without RCM (corresponding to a NNE of 19.41), the cost corresponded to €2932 per melanoma removed. Introduction of RCM into the standard procedure resulted in a mean total cost per patient of €105, or a cost of €2133 per melanoma removed, using a NNE of 6.25 after RCM and an estimate of 14% of lesions referred to follow-up excised later because of changes on digital dermoscopy monitoring.

Overall, the use of RCM in addition to dermoscopy for the diagnosis of skin lesions suspicious of skin cancer led to a large reduction in the number of benign lesions excised and a reduction in the mean total cost per patient from the point of lesion assessment to the point of excision or discharge.

One limitation of the study was its retrospective design. Another limitation was the omission of training costs associated with the use of RCM. Clinicians assessing skin lesions with RCM should achieve adequate expertise in image reading. According to the authors, a minimum of 6 months’ full-time training, including the evaluation of more than 4000 cases, is required in order to obtain adequate levels of diagnostic accuracy and confidence. However, the authors noted that the development of a teledermatology confocal-dedicated platform might reduce the time needed to effectively implement and exploit the technique into the clinical practice, through distant teaching and diagnostic support to new centres and users.

The study suggests that the use of RCM in addition to dermoscopy for the diagnosis of skin cancer may represent a cost-effective strategy as it leads to better outcomes and cost-savings to the health-care system. It should be noted, however, that, as the study was conducted in Italy, its findings may not be generalisable to the UK setting, as there may be differences between the two health-care settings in terms of prevalence of the various skin cancer types, the population phenotype distribution, the clinical pathways for the diagnosis, assessment and management of skin cancer, the level of experience of clinicians in the use of RCM and relevant unit costs.

The methodological quality of the study by Pellacani et al. 46 assessed against the NICE reference checklist for economic evaluations, is presented in Table 19. The evidence table with the summary of methods and results of the study is provided in Table 20.

Resource-use and costing studies

A total of 277 papers were identified from the systematic search of the literature. Of those, 205 were excluded on the basis of title and abstract and 63 were duplicates. Therefore, a total of nine papers were identified as potentially relevant and were ordered for full review based on the criteria listed in Box 2. On the basis of the full text, six studies were excluded. Reasons for exclusion of the ordered papers are provided in Appendix 5. The remaining three studies identified from the search included relevant UK cost data on skin cancer.

Figure 4 provides the flow chart of the process of the systematic search for resource-use and costing studies. 48–50

FIGURE 4.

Flow chart of the process of the systematic search for resource-use and costing studies.

Of the three studies included in this review, one48 was an economic evaluation of a diagnostic aid (the MoleMate system; MedX Health Corp., Mississauga, ON, Canada) versus best practice in people with pigmented skin lesions in primary care. The other two studies49,50 estimated the cost of skin cancer in England.

Wilson et al. 48 conducted a model-based economic evaluation that assessed the lifetime costs and QALYs associated with the diagnostic assessment of people with at least one suspicious pigmented skin lesion presenting to UK primary care. The economic model consisted of a decision tree and a Markov model that followed TP, TN, FP and FN cases (based on diagnostic assessment) over a lifetime. The analysis, which adopted the NHS perspective, considered explicitly only the costs and outcomes of melanoma, as it did not differentiate between melanoma and non-melanoma skin cancer. Costs included diagnostic assessment costs and costs of TP, TN, FP and FN cases over a lifetime. Costs were calculated using a bottom-up approach. Treatment costs were estimated according to stage of melanoma, including initial treatment (biopsy excision and definitive surgery), investigations, follow-up surgery for positive lymph nodes, treatment of metastatic disease, follow-up and terminal care. Resource use and costs associated with the management of each melanoma stage were reported separately. Resource-use estimates for the treatment of distinct melanoma stages were based on the 2010 UK guidelines for the management of cutaneous melanoma,13 supplemented by expert opinion. Unit costs were based on the NHS Reference Costs 2008 to 2009. 51 The cost year was 2009.

The study appears to report cost data that are potentially useful for economic modelling. However, clinical experts advised that costs associated with the treatment of more advanced melanoma stages (stages III and IV) are likely to have changed recently, with the introduction of new chemotherapeutic agents, such as ipilimumab (YERVOY^®, Bristol-Myers Squibb) and vemurafenib (Zelboraf^®, Roche Products Ltd), in the treatment of advanced melanoma in the NHS.

Morris et al. 49 reported the costs associated with malignant melanoma and other malignant neoplasms of the skin in England from a societal perspective. Health-care costs included GP assessment, inpatient stays, outpatient attendances and day cases; in addition, travel costs, incapacity benefits and productivity losses were estimated. The cost year was 2002. Costs were estimated using a top-down approach; total costs were divided by the number of registrations to estimate the mean cost per registration. Resource-use data and unit costs were taken from national sources. The study reported the mean NHS and societal cost per registration of melanoma to be £2179 and £20,020, respectively. The mean NHS and societal cost per registration of other malignant skin neoplasms was £1149 and £1413, respectively.

The resource-use data utilised by this study in order to estimate costs are out of date, as some estimates are > 20 years old. Moreover, the top-down approach allows only a rough estimation of relevant costs. Finally, it is noted that the study provides an overall cost per case with skin cancer (either melanoma or non-melanoma) but, in the case of melanoma, does not report costs by stage of skin cancer.

Vallejo-Torres et al. 50 also reported the costs associated with melanoma and non-melanoma skin cancer in England, but from a NHS perspective. The study used both a top-down and a bottom-up approach in order to produce cost estimates. The cost year was 2008. The top-down approach was adapted from Morris et al. 49 using more up-to-date costs, and was not used to estimate a cost per case. The bottom-up approach used a simplified model of skin cancer care in the NHS, which utilised probabilities of people with suspected skin cancer using different treatment pathways; costs for each pathway were estimated separately. Data to populate the model were taken from UK guidelines for the management of skin cancer,13,14 other published reports and clinical expert opinion. Treatment pathways included initial examination, treatment in primary care or referral to a specialist, diagnostic biopsy of suspicious lesions and treatment according to the biopsy results.

Even though the study by Vallejo-Torres et al. 50 uses more up-to-date resource-use figures and unit costs, the probabilities of treatment received by patients may no longer represent clinical practice, as they were based on an outdated study. 52 Moreover, the study provides separately costs per treatment pathway, but, in the case of melanoma, does not report costs by stage of skin cancer.

The overall methods of the resource-use and costing studies, the resource-use elements that are potentially relevant for the economic model developed for this report and the estimated costs associated with management of skin cancer are presented in Table 21.

Studies reporting utility data

A total of 11,497 citations were identified from the systematic literature search. Of those, 3547 were duplicates and 7909 studies were excluded on the basis of title and abstract. A total of 43 full texts were assessed against the inclusion criteria listed in Box 2; these included 41 studies identified from the database search and the two studies identified from the reference list search.

Of the 41 ordered studies identified from the database search, 17 were cost-effectiveness studies that obtained utility values from the literature to estimate QALYs. Consequently, the sources used to inform the utility values in these studies were identified and reviewed for inclusion. Two further studies were identified from the references lists of those 17 cost-effectiveness studies retrieved from the database search. A full list of the sources used to inform the cost-effectiveness studies is provided in Appendix 5. After full-text review, 38 studies were excluded. Reasons for exclusion of the ordered papers are also presented in Appendix 5. Out of the 43 full texts assessed for inclusion, a total of five studies met the inclusion criteria defined in Box 2. 53–57

After the systematic search was completed, the EAG was informed by experts in the field of an additional recently published paper that provided relevant utility data. 58 Subsequently, the EAG included this paper in the systematic review, after assessing the full text against the set inclusion criteria. It should be noted that one of the inclusion criteria specifies a requirement for a choice-based technique for valuation [i.e. time trade-off (TTO) or standard gamble (SG)]; Tromme et al. 58 does not meet this criterion, as valuation of health states was based on the visual analogue scale (VAS). 58 However, as this study reported utility values that were generated using the EQ-5D, which is the preferred measure by NICE, and the EQ-5D has been valued by the Flemish population in Belgium using the VAS, it was decided to relax the inclusion criteria in order to include this study in this review. None of the 38 potentially relevant studies that had been excluded according to the predefined inclusion criteria met the relaxed inclusion criteria.

In total, six studies were included in the review of studies reporting preference-based HRQoL data (utility data) for skin cancer. 53–58

Figure 5 provides the flow chart of the process of the systematic search for studies reporting utility data.

FIGURE 5.

Flow chart of the process of the systematic search for studies reporting utility data. HTA, Health Technology Assessment; NHS EED, NHS Economic Evaluation Database.

Four of the six studies included in the review reported utility values for melanoma-related health states. 53–55,58 One study reported utility values for health states of advanced BCC56 and the other study reported utility values associated with scarring following facial and auricular non-melanoma skin cancer surgery and reconstruction. 57

Askew et al. 53 reported EQ-5D utility values for different melanoma stages derived from 273 patients with melanoma, 75 of whom were undergoing treatment and 198 were undergoing follow-up surveillance at the time of the study; all attended a tertiary cancer care centre in the USA. The median age of the study sample was 52 years, 98% were white and 58% were male. The number of patients at each melanoma stage was 102 at stage I/II, 100 at stage III and 71 at stage IV. The utility values were generated using patient responses on the EQ-5D. The US EQ-5D tariff was used, which has been developed following a valuation survey of 4048 representative members of the US population using TTO. 59

Beusterien et al. 54 reported utility values for various hypothetical advanced melanoma-related health states elicited from 140 members of the general population (77 from Australia and 63 from the UK), using SG. The hypothetical health states (vignettes) included four advanced melanoma treatment-related response states, one symptomatic melanoma state and nine toxicity-related health states. The hypothetical health states were constructed based on published literature and refined following an iterative review by five clinical experts, two oncology nurses, three quality-of-life researchers and a pilot test with individuals from the general public. The four treatment-related response states were defined as follows: partial response was defined by a > 50% decrease in lesion mass; stable disease was defined by a > 25% decrease or increase in lesion mass; progressive disease was defined by the appearance of new lesions or a > 25% increase in lesion mass; and for best supportive care there was no indicated or desired cancer treatment. A symptomatic melanoma health state represented symptoms experienced in advanced melanoma. The health states were described as being treated for cancer (melanoma was not specified), whether or not treatment is working, and changes in tumour size, pain levels, appetite, effort required to perform daily activities and fatigue. Each of the toxicity descriptions was described in association with partial response so that the utility decrements for toxicities could be calculated by subtracting the utility for partial response from the utility of the toxicity state.

King et al. 55 developed vignettes describing health states associated with each of the melanoma stages (I–IV) based on published literature and relevant websites; the hypothetical health states were valued by 163 adult patients with melanoma attending a cancer clinic in the USA using TTO. Patients were divided into new cases (if they were ≤ 1 year from diagnosis) and established cases (if they were > 1 year after diagnosis or > 6 months if stage IV). Patients were asked to value stages other than their own: patients with stage I disease imagined having a new diagnosis of stage II, III or IV disease, while patients with higher-stage disease imagined the impact of a new stage I diagnosis. Utilities derived from new cases, established cases and all patients participating in the study were reported separately.

Tromme et al. 58 reported EQ-5D utility values for different melanoma stages derived from 356 patients with melanoma. Patients completed the five-level version of the EQ-5D (EQ-5D-5L); 39 patients completed the EQ-5D-5L questionnaire twice, as they were seen in two different phases (treatment and follow-up) and/or stages during the study. Patients were classified into eight groups using four melanoma stages (I, II, III and IV), with each stage subdivided into treatment and remission phases.

Patients with stage 0 and Ia melanoma were pooled with the justification that they had marginal differences regarding their surgical treatment and follow-up. Patients with stage Ib and II melanoma were also pooled because, according to the authors, these patients had undergone sentinel lymph node biopsy (SLNB) that had not been followed by elective node dissection and also because of evidence that surgical resection margins did not appear to influence HRQoL.

Based on expert opinion, treatment duration was estimated to be 1, 2 and 3 months for stages 0–Ia, Ib–II and III, respectively, and > 10 months for stage IV. The remission period for stages 0/Ia and Ib/II was estimated at 2 years of follow-up, as it has been shown that after 2 years the HRQoL of these patients is similar to that of the general population. 60 Patients with stage IV melanoma in remission but still under treatment were classified as patients under treatment in order for the impact of side effects on HRQoL to be captured. The mean age of the patients was 52.6 years and 74% were male. EQ-5D-5L profiles were first mapped onto EQ-5D three levels (EQ-5D-3L) profiles, which were subsequently converted into utility values using the Belgian EQ-5D tariff, which has been developed following a valuation survey of 2754 Flemish adults from the general public in Belgium using the VAS. 61

Shingler et al. 56 reported utility values for a number of hypothetical advanced BCC-related health states elicited from a representative sample of 100 members of the UK general public using TTO. The health state vignettes associated with advanced BCC were constructed based on a literature review, consultation with two clinical experts and validation/piloting with three members of the general public. At the end of this process, nine health state vignettes were developed, reflecting level of treatment response. The nine vignettes describing advanced BCC health states were as follows: complete response; post-surgical state; partial response with small (2 cm) or large (6 cm) growth; stable disease with small (2 cm) or large (6 cm) growth or multiple growths (2 cm); and progressed disease with small (2 cm) or large (6 cm) growth.

Seidler et al. 57 developed simple health state vignettes describing the type of repair and subsequent scar after facial and auricular non-melanoma skin cancer surgery and reconstruction. 57 One state comprised surgery for facial non-melanoma skin cancer, a second state described simple repairs or scars (granulation and primary closure) as a result of surgery and a third state described complex repairs or scars (local flap and graft) because of surgery of non-melanoma skin cancer. Five healthy adults from the general public in the USA valued the three health states using TTO.

Table 22 summarises the methods used to derive and value health states associated with skin cancer and the resulting utility scores, as reported in the six studies included in this systematic review.

According to NICE guidance47 on the selection of utility values for use in cost–utility analysis, the measurement of changes in HRQoL should be reported directly by people with the condition examined, and the valuation of health states should be based on public preferences elicited using a choice-based method, such as the TTO or SG, in a representative sample of the UK population. When changes in HRQoL cannot be obtained directly from the people with the condition examined, then data should be obtained from their carers. NICE recommends EQ-5D for use in cost–utility analyses of interventions for adults. When EQ-5D scores are not available or are inappropriate for the condition or effects of treatment, NICE recommends that the valuation methods be fully described and comparable to those used for the EQ-5D. 47

None of the studies included in the review meet the above criteria set by NICE. Two of the studies53,58 used the EQ-5D for the description of HRQoL experienced by patients with melanoma. However, none of them used the UK EQ-5D tariff62 for the valuation of health states; Askew et al. 53 used the USA EQ-5D tariff, which was developed using TTO, whereas Tromme et al. 58 used the Belgian EQ-5D tariff, which was developed using the VAS – a valuation method that is not choice based and thus is not among NICE-preferred valuation methods.

All the remaining studies generated utility values for health states described in vignettes. Of those, Beusterien et al. 54 elicited utility values for melanoma-related health states from members of the general population in Australia but also in the UK, so, in this aspect, the study meets the NICE criterion for valuation of states by the UK general population. The same applies to Shingler et al. ,56 who reported utility values for advanced BCC-related health states obtained from members of the UK general public. In contrast, King et al. 55 reported melanoma-related utility values elicited from patients with melanoma in the USA, while Seidler et al. 57 reported utility values associated with facial non-melanoma skin cancer surgery and reconstruction that were elicited from only five healthy adults in the USA.

A comparison of the utility values available for melanoma-related health states according to stage revealed that the utility values reported by Askew et al. 53 for melanoma stages III and IV are considerably higher than those reported by King et al. 55 and Tromme et al. 58 Moreover, the utility values reported by Tromme et al. 58 for melanoma early stages I and II are substantially lower than the utility values reported for the same stages in Askew et al. 53 and King et al. 55 These discrepancies are potentially attributable to differences in measurement and valuation across the three studies. Measurement of HRQoL in two studies53,58 was taken from patients with melanoma using the EQ-5D, whereas King et al. 55 used vignettes to describe the HRQoL associated with melanoma stages. In the two studies reporting EQ-5D-based utility values,53,58 values had been elicited from members of the general population in two different countries (USA and Belgium) using two different valuation techniques (TTO and VAS). King et al. 55 elicited values from patients with melanoma. 55 These differences in valuation may also be responsible for the differences in resulting utility values.

In addition, it is noted that utility values for stage III appear to be lower than utility values for stage IV in Tromme et al. ;58 however, this may be attributable to the variation in values because of the small number of patients providing EQ-5D ratings for stage III in treatment (n = 15) and stage IV in remission (n = 14).

The available utility values for skin cancer, the methods used in their development and underlying limitations as well as their eligibility for use in economic modelling according to NICE criteria are discussed in the appropriate subsections of Economic modelling.

Overall objective

The overall objective of the economic analysis, as defined by the scope of this diagnostic assessment, was to assess the cost-effectiveness of the VivaScope 1500 and 3000 imaging systems in the diagnosis of potentially malignant skin lesions, including LM, and the margin delineation of diagnosed malignant lesions, including LM, prior to surgical treatment. Not all potentially malignant or diagnosed skin lesions are suitable for diagnosis or presurgical margin delineation with the use of the VivaScope imaging system. The selection of population groups with suspected (or diagnosed) skin cancer for consideration in economic modelling was determined by the availability of relevant clinical data and clinical expert opinion.

Study population

The VivaScope 1500 and 3000 imaging system was assessed in the diagnosis of skin cancer in the following populations:

• people with suspected melanoma who have equivocal lesions following dermoscopy

• people with suspected BCC whose lesions have an equivocal or positive result on dermoscopy, to make the diagnosis or to confirm diagnosis, respectively, as an alternative to diagnostic biopsy.

The above populations were considered to be the most relevant to undergo diagnostic assessment with VivaScope, according to clinical experts to the EAG. Clinical experts also advised that they form an opinion about the type of skin cancer they will find prior to examination with VivaScope, and hence the type of suspected skin cancer was prespecified at early stages of designing the economic model. The NICE scope63 defines the study population as ‘people with equivocal lesions following dermoscopy’; however, clinical experts advised the EAG that the use of VivaScope in suspected BCC lesions has two purposes: (1) to make a diagnosis when results of dermoscopy are not certain; and (2) to confirm diagnosis in lesions that are found positive on dermoscopy. In both cases the VivaScope is used as an alternative to diagnostic biopsy. Thus, the economic model considered all people with suspected BCC lesions eligible for dermoscopy, and not only those with equivocal lesions suspected to be BCC, as the latter are rather a minority of the cases eligible for examination with VivaScope.

Equivocal lesions among those suspected of being melanoma include any suspected melanoma lesions based on a number of characteristics on dermoscopy, with the exception of clearly positive (malignant) lesions that have all the dermoscopic characteristics of melanoma and clearly negative (benign) lesions that show no features of melanoma (no changes) on dermoscopy. The risk of equivocal lesions being malignant is overall low. There are different degrees of ‘equivocalness’, depending on the dermoscopic characteristics of the lesion and subjective experience and interpretation. Clinical expert advice indicated that highly suspicious equivocal lesions are lesions with at least two positive dermoscopic features including one major criterion or three minor positive features suggestive of melanoma, and/or lesions clearly changed after digital follow-up, and/or new or growing lesions in an adult with at least one dermoscopic positive criterion, or papular/nodular or pink or spitzoid lesions. In all of those cases, excision is prompted and examination with VivaScope does not represent a real advantage, as the risk of missing a melanoma remains too high. Moderately or low suspicious equivocal lesions are lesions with only one major dermoscopic positive feature or two minor features, and/or no clear history or minor changes. In such cases, excision is possible but other options could be taken into account, such as digital follow-up, especially in the case of flat lesions in patients with multiple moles; however, digital follow-up can delay a melanoma diagnosis. The majority of moderately or low suspicious equivocal lesions that are excised are benign, and examination with VivaScope can play a major role in reducing this burden of unnecessary excisions.

Clinical experts advised that VivaScope is less suitable for the detection and assessment of skin lesions suspicious of SCC, as this type of skin cancer is usually scaly because of severe hyperkeratosis. This often limits the evaluation of SCC lesions, as it is more difficult to capture images of structures deeper in the tissue. Moreover, no evidence on the diagnostic accuracy of VivaScope in this type of skin cancer was identified in the systematic review of clinical evidence. Therefore, it was decided not to include people with skin lesions suspicious of SCC in the diagnostic economic model.

Regarding margin delineation, VivaScope 3000 was assessed in the following population:

• patients with LM prior to surgical management.

According to clinical expert advice, margin delineation of melanomas with VivaScope is not useful in clinical practice, as the margins of melanomas are clearly defined and can be completely excised following BAD guidance;13 consequently, VivaScope mapping of melanomas does not offer any clinical utility and, therefore, was not considered further for economic modelling.

Clinical experts advised that margin delineation of BCCs using VivaScope may be difficult, as BCCs may be too deep for their margins to be accurately mapped with VivaScope. Therefore, it was decided not to consider margin delineation of BCC lesions with the use of VivaScope in the economic model, also considering the lack of evidence in this area. Nevertheless, it is acknowledged that, although margin delineation of BCCs using VivaScope prior to surgical excision was not considered in the economic analysis, this may be used as an alternative to Mohs surgery, as advised by clinical experts.

VivaScope is not appropriate for the assessment of SCC lesion margins due to the scaly nature of SCC making it difficult to view using imaging techniques.

Economic models developed: decision problems addressed

According to the study populations that were identified as relevant for the economic evaluation of VivaScope, three separate ‘part’ economic models were developed:

1. Use of VivaScope in the diagnosis of equivocal lesions suspected of being melanoma. This model assessed the cost-effectiveness of VivaScope 1500 and 3000, as one integrated system, assuming that both devices will be available for the diagnosis of equivocal lesions but each will be used as appropriate according to the location of the equivocal lesion to be examined.

2. Use of VivaScope in the diagnosis of suspected BCC lesions following a positive or equivocal finding on dermoscopy. As with the previous model, this model assessed the cost-effectiveness of VivaScope 1500 and 3000, as one integrated system, assuming that both devices will be available for the diagnosis of suspected BCC lesions but each will be used as appropriate according to the location of the skin lesion to be examined.

3. Use of VivaScope for the margin delineation of LM prior to surgical therapy. This model assessed the cost-effectiveness of VivaScope 3000 as a stand-alone device, as only this device is appropriate for margin delineation.

Development of two separate models, one for the diagnosis of equivocal lesions suspected of being melanoma and one for suspected BCC lesions with a positive or equivocal dermoscopy finding, was necessary because both the diagnostic accuracy of VivaScope and the treatment pathways and associated costs and outcomes following diagnosis vary greatly between these two types of skin cancer.

Using the results of the above three ‘part’ models, five economic analyses were undertaken, examining the cost-effectiveness of VivaScope in:

1. the diagnostic assessment of equivocal lesions suspicious of melanoma (integrated use of VivaScope 1500 and 3000)

2. the diagnostic assessment of lesions suspicious of BCC following a positive or equivocal result on dermoscopy (integrated use of VivaScope 1500 and 3000)

3. the diagnostic assessment of skin lesions suspicious of skin cancer, either melanoma (following an equivocal finding on dermoscopy) or BCC (following a positive or equivocal finding on dermoscopy) – this analysis combined the results of the two ‘part’ models.

4. the margin delineation of LM prior to surgical treatment (use of VivaScope 3000 as a stand-alone device)

5. the diagnostic assessment of skin lesions suspected of being either melanoma or BCC, and the margin delineation of LM (integrated use of VivaScope 1500 and 3000) – this analysis combined the results of all three ‘part’ models.

The final economic analysis synthesised all cost and effectiveness data from each of the ‘part’ economic models to obtain an estimate of the overall cost-effectiveness of the VivaScope imaging system used for all indicated purposes assessed in economic modelling in a skin cancer multidisciplinary team (MDT) service.

The analyses that combined results of ‘part’ models used weighted total costs and benefits according to the expected relative number of each type of lesion diagnosed and/or mapped with VivaScope in one dermatology MDT service.

Costs and outcomes considered in the analysis

The perspective of the analysis was that of the NHS and Personal Social Services (PSS). Costs consisted of the costs of the intervention, that is VivaScope (including equipment and maintenance costs, costs of consumables, staff training and staff time required for the examination), the costs associated with the comparators used in the the analysis (such as the costs of biopsy, histological examination and monitoring including any required consultations with clinicians), the costs of management of skin lesions following correct (i.e. TN and TP cases) or incorrect (FN and FP cases) diagnosis, as well as the costs incurred following the presurgical mapping of malignant skin lesions. The costs of management of future events such as progression and recurrence of skin cancer, where relevant, were also considered. All costs were expressed in 2014 prices.

The outcome measure of the economic analysis was the QALY. Discounting of costs and outcomes was applied at an annual rate of 3.5%, in accordance with NICE methodology guidance. 47

Sources of model input parameters

The clinical effectiveness parameters required for the economic models (diagnostic accuracy of VivaScope 1500 and 3000 and clinical outcomes on margin delineation) were informed, where possible, by the systematic review of the clinical evidence reported in Chapter 3. A non-systematic review of model-based economic studies assessing strategies and interventions for the prevention, assessment or management of skin cancer was also undertaken, aiming to provide an insight into the modelling methods in the area of skin cancer and also identify relevant input parameters that could be potentially utilised in the economic models assessing VivaScope. These studies were predominantly identified by re-running the search for existing economic evaluations of VivaScope (described in Systematic literature review of existing economic evidence, Methods) after omitting the terms capturing the interventions and comparators of interest from the search strategy. The search resulted in a very high number of hits (approximately 9000) that did not permit a review of the findings in a systematic way because of time and resource constraints. Nevertheless, this review helped identify a range of useful clinical (as well as resource-use) data and model structural components that contributed to the construction of the model structures for the economic assessment of VivaScope. In addition, relevant NICE guidance (including clinical and public health guidelines, technology appraisals and interventional procedure guidance) was reviewed for clinical and cost data that could be potentially useful in economic modelling.

Preference-based data on the HRQoL of people experiencing health states or events associated with suspected or confirmed skin cancer were derived from the relevant published literature identified in the systematic review, the results of which are provided in Systematic literature review of existing economic evidence, Results.

Following clinical expert advice, the EAG undertook a review of conference abstracts presented at the British Association of Dermatologists’ annual meetings since 2010, which are available from the British Journal of Dermatology. This review aimed to identify audits reporting data on health service use from patients with skin cancer in the UK, as well as recent trends in epidemiological data directly relevant to the UK population that could inform the economic models. Clinical experts also provided references to studies reporting data that were potentially useful in populating the economic models.

Finally, at all steps of designing the economic models, clinical expert opinion was sought to confirm that diagnostic and assessment pathways were consistent with current clinical practice in the UK, as well as with anticipated changes in practice following a potential introduction of VivaScope within the NHS context. Clinical expert opinion was also employed to supplement the economic models with parameter estimates, in areas where relevant published evidence was lacking.

The costs associated with the intervention (VivaScope 1500 and 3000 imaging systems), including the purchase price of the equipment and parts and maintenance costs, were provided by the company. Other health-care unit costs were obtained from national sources such as the NHS drug tariff for February 2015,64 the national Unit Costs of Health and Social Care 201465 and the NHS Reference Costs 2013 to 2014 for 2014. 12 The NHS reference costs were preferred over the Payment by Results tariffs because they represent actual national average costs incurred as a result of health-care services provided by the NHS, and hence they reflect opportunity costs, whereas the Payment by Results tariffs represent payments rather than the actual cost of services to the NHS.

Annual number of cases eligible for examination with VivaScope in a dermatology multidisciplinary team clinic in the UK

The annual number of cases eligible for examination with VivaScope in a dermatology MDT clinic was needed in order to determine the total cost per case associated with a VivaScope examination, as the overall cost of VivaScope (including purchase and maintenance cost, training costs and any other ancillary costs) is spread across the number of lesions examined. Given the high cost of purchasing VivaScope and the considerable training required for obtaining and interpreting VivaScope images, an adequate number of VivaScope examinations needs to be performed every year, so that the benefit from VivaScope use offsets the intervention cost.

In order to estimate the total number of people that are assessed with VivaScope in 1 year, three approaches were followed.

The first approach was to ask clinical experts working in the dermatology department of Guy’s and St Thomas’ Hospital, London, UK, where VivaScope is currently in use, about the annual number of cases examined with VivaScope in their practice.

This approach yielded the following information:

• Approximately one suspected melanoma is assessed with VivaScope per week; however, it was suggested that this number is probably lower than the typical number of lesions suspicious of melanoma that would normally be examined by a tertiary service and that would be eligible for examination with VivaScope.

• Approximately 15 suspected BCC lesions are assessed with VivaScope per week; however, it was suggested that this number might be higher than the typical number of suspected BCC lesions that would normally be examined by a tertiary dermatology service.

• Approximately one or two LMs are mapped with VivaScope per week, but this includes LMs planned for surgical therapy as well as radiotherapy and topical immunotherapy.

Based on the above information, the annual number of lesions examined with VivaScope was estimated to comprise 75–100 equivocal lesions suspicious of melanoma (estimated as 1.5–2 expected to be seen per week × 50 weeks); 500 suspected BCC lesions (estimated as 10 expected to be seen per week × 50 weeks); and 75 LMs prior to treatment (estimated as 1.5 expected to be seen per week × 50 weeks, and considering that the vast majority of LMs are treated surgically, as advised by clinical experts).

The second approach was to seek information from clinical experts working in other dermatology services, who were approached for expert opinion and advice on the preparation of this report, and on the annual number of suspected melanomas, suspected BCCs and LMs eligible for examination with VivaScope that were assessed in their practice.

This approach yielded the following information:

• The dermatology clinic at the Norfolk and Norwich University Hospital examines approximately 600–800 lesions suspicious of melanoma per year. No information was available on the number of lesions suspicious of BCC or LMs examined per year.

• The dermatology service at the Lincoln hospitals serves a population of about 0.75 million. Using population incidence data, it was estimated that every year the service diagnoses about 160 melanomas, 1000 BCCs and roughly 60–80 LMs. The vast majority of BCCs are easy to diagnose clinically or with dermoscopy; sometimes they are so typical that no dermoscopy is needed.

• The Department of Dermatology at the Chelsea and Westminster Hospital examines around 300 suspected BCCs and manages at most 20 LMs annually. No information was available on the number of suspected melanomas examined per year.

Clinical experts advised that, of every five or six lesions that are excised because of suspicion of melanoma, one melanoma is confirmed. Using the estimate of 160 diagnosed melanomas, the number of suspected melanomas examined by the service in Lincoln (i.e. lesions giving a positive or equivocal result on dermoscopy) should be approximately 800–960 per year.

Of the suspected melanomas examined in each service, only those giving an equivocal finding on dermoscopy would be eligible for examination with VivaScope. Therefore, to estimate the number of suspected melanomas eligible for examination with VivaScope in each service, the proportion of equivocal lesions among the number of suspected melanomas examined in each service is needed. For this reason, a review of the studies included in the systematic review of clinical evidence reported in Chapter 3 was undertaken, attempting to identify the proportion of suspected melanomas examined by a dermatology MDT clinic that give an equivocal finding on dermoscopy. The review considered studies reporting prospective or retrospective recruitment of consecutive people attending a dermatology clinic for skin lesions suspicious of melanoma, which were assessed with a dermoscope, as they were likely to be more representative of the population of people with suspected melanomas likely to be seen at a dermatology clinic. Studies that had selectively recruited people with suspicious skin lesions and those that assessed retrospectively lesions that had already been excised on the basis of their dermoscopic features were not considered, as their study samples were not necessarily representative of the study population. In addition, studies that had excluded ‘clear-cut positive lesions on dermoscopy’ from recruitment were not considered useful, as they provided an overestimate of the proportion of equivocal lesions among the total number of lesions examined with dermoscopy.

The only suitable study included in the systematic review of clinical evidence reported in Chapter 3 was that by Alarcon et al. ,30 who assessed the impact of a VivaScope examination of equivocal lesions suspicious of melanoma following diagnostic assessment with dermoscopy. From 5520 patients attending a hospital dermatology unit in Barcelona, the study identified 1534 people with lesions suspicious of melanoma who underwent dermoscopy. In 1191 of these lesions, the finding, according to the authors, was clear and the lesions were scheduled either for immediate excision or for digital follow-up. In the remaining 343 lesions, the findings were equivocal and thus these lesions were suitable for examination with VivaScope 1500. Thus, the percentage of equivocal lesions among all lesions suspicious of melanoma and assessed with dermoscopy was 22.4% (343/1534).

A Belgian observational study assessed the extent of cost reduction resulting from the use of sequential digital dermoscopy in people presenting to dermatologists because of concern about melanoma and having 1–3 equivocal melanocytic lesions. 66 The study reported that, of the 9360 consecutive people with 1–3 lesions suspicious of cancer that were assessed with a dermoscope over 1 year (2009–10), 822 had equivocal lesions, according to dermatologists, making the percentage of equivocal lesions among lesions suspicious of melanoma 8.78%. However, the study population was people presenting to dermatology services rather than being referred, and, therefore, the prevalence of melanoma, and subsequently the prevalence of equivocal lesions, was most likely lower than the prevalence of melanoma and prevalence of equivocal lesions in populations referred to dermatology MDTs from primary care in the UK.

It should be noted that the proportion of equivocal lesions among lesions suspicious of melanoma that are examined with a dermoscope is affected by a number of other factors, such as the experience of the dermatologist performing the examination and the underlying prevalence of melanomas.

Expert opinion indicated that the proportion of equivocal lesions out of lesions suspicious of melanoma undergoing dermoscopic evaluation in England must be between the figures observed in the two studies described above. 30,66 Therefore, estimations of the number of lesions suspicious of melanoma that are suitable for a VivaScope examination (i.e. they have an equivocal finding on dermoscopy) were based on the assumption that approximately 15% of lesions examined by skin cancer MDT services are equivocal.

Using the estimates of suspected melanomas seen annually in each of the two services (600–800 cases at the dermatology clinic at the Norfolk and Norwich University Hospital and 800–960 cases at the dermatology clinic in Lincoln) and a proportion of equivocal lesions following dermoscopy of 15% among all cases examined, the estimated number of equivocal lesions suspicious of melanoma that would be eligible for examination with VivaScope seen by each service per year was 90–120 in Norfolk and Norwich University Hospital and 120–144 in Lincoln.

Regarding BCCs diagnosed at the dermatology clinic at Lincoln Hospitals, the experts considered that the vast majority are easy to diagnose, sometimes without the use of a dermoscope. Expert opinion suggested that, out of 1000 confirmed BCCs, in 600–700 clinical examination would reveal a clear-cut picture (and a positive finding on dermoscopy). The remaining 300–400 confirmed BCC lesions would be identified after roughly 500–600 lesions suspicious of BCC had examined with VivaScope (following an equivocal finding on dermoscopy). In total, at least 500–600 lesions per year would be eligible for examination with VivaScope to make a diagnosis, and 600–700 clear-cut positive BCC lesions would be eligible for examination with VivaScope for confirmation of diagnosis, leading to a total estimate of 1200–1400 lesions suspicious of BCC that would be eligible for examination with VivaScope in the dermatology service of Lincoln hospitals annually.

The number of suspected BCCs examined at the Department of Dermatology of the Chelsea and Westminster Hospital annually is approximately 300. Similar to the above estimations, about 180–200 would be expected to be positive on dermoscopy and 100–120 equivocal; the latter would be identified after roughly 150–200 lesions suspicious of BCC have been examined with VivaScope following an equivocal finding on dermoscopy, resulting in a total estimate of around 330–400 lesions suspicious of BCC that would be eligible for examination with VivaScope in the Department of Dermatology of the Chelsea and Westminster Hospital.

The number of LMs examined annually at the Department of Dermatology of the Chelsea and Westminster Hospital was 20 at maximum, whereas at Lincoln hospitals roughly 60–80 LMs are diagnosed every year. It was assumed that practically all of them are treated surgically.

The third approach was to estimate the numbers of lesions/cases eligible for VivaScope examination indirectly, by estimating the numbers of people being referred to dermatology services each year in the UK and, from that, estimate the number of people in a large dermatology MDT service.

This approach considered dermatology MDT clinics serving a large population of people, which are likely to see a high number of skin lesions suspicious of skin cancer. SSMDT clinics were selected for this purpose, as it is expected that more suspected skin cancers are referred here than in LSMDTs. These serve a catchment population for referral (their own local catchment plus the catchment of referring LSMDTs) of at least 750,000 and serve as the LSMDT for their local (secondary) catchment population. 67

In 2009–10, 882,000 patients were referred to dermatologists in England (approximately 16 per 1000 population). 68 Up to 50% of referrals relate to skin cancer (including both diagnosis and management). Dermatologists screen > 90% of skin cancer referrals and treat approximately 75%. 68 In the period between 2000 and 2007, there was an increase of about 5.6% in new patients visiting dermatology specialists. 69 This is an increase of approximately 0.8% per year. Applying this annual rate of increase to the data from 2010, in 2015 the expected number of people referred to dermatologists in England is 16.63 per 1000 population. With 50% of referrals relating to skin cancer and 90% of them being screened, this results in an estimate of approximately 7.49 examinations for skin cancer per 1000 population per year.

Assuming a catchment area of 750,000, one SSMDT would examine around 5614 people for skin cancer per year. Assuming that the ratio of referred lesions suspicious of BCC, SCC and melanoma is approximately the same as the ratio of confirmed skin cancers, and taking into account the fact that in 2011 there were 102,628 cases of non-melanoma skin cancer in the UK (of which BCCs make up 74%) and 13,348 cases of melanoma in the UK,2 then, of the 5614 people examined for skin cancer annually, 11.5% (646) would be examined for suspected melanoma and 65.5% (3676) would be examined for suspected BCC.

Using an estimated proportion of equivocal lesions among lesions suspicious of melanoma of 15%, a SSMDT serving a population of 750,000 would see 97 equivocal lesions suspicious of melanoma per year.

Clinical expert advice was that the number of 3676 lesions suspicious of BCC appears to be unrealistically high. Therefore, it was assumed that only 50% of them were actually suspected to be BCCs. Using estimates described earlier, it was calculated that roughly 2000–2300 lesions suspicious of BCC would be potentially eligible for a VivaScope examination.

Epidemiological data specific to LM are rather sparse and not routinely available in UK cancer statistics. Incidence data on LM were reported in a US study that identified all adult residents with a first lifetime diagnosis of LM between 1970 and 2007 in Olmsted County, MN, USA. 70 The study reported that the overall age- and sex-adjusted incidence of LM among adults was 6.3 per 100,000 person-years, increasing from 2.2 between 1970 and 1989 to 13.7 between 2004 and 2007. 70 Although the incidence of LM in the UK population, as well as the mixture of the UK population, may be different from that in Minnesota, using the incidence of 13.7 per 100,000 person-years in a population of 750,000 people would result in 103 cases identified and treated in a dermatology service annually.

The estimates derived from the three approaches are summarised in Table 23.

As can be seen, with the exception of BCC lesions, for which the range of estimates is wide, the estimated number of equivocal lesions suspicious of melanoma and the estimated number of LMs that are eligible for examination with VivaScope using each of the three approaches are very close. In order to estimate the cost of VivaScope per skin lesion assessed, the following estimates in the number of lesions examined annually with VivaScope in a dermatology service were utilised:

• 100 equivocal lesions suspicious of melanoma

• 500 lesions suspicious of BCC (use of a low, relatively conservative estimate, which was based on information derived from the only setting in the UK that currently uses VivaScope for the diagnostic or presurgical assessment of skin lesions)

• 75 LMs prior to treatment.

Cost of VivaScope 1500 and 3000

This section reports the costs associated with examination of skin lesions with the VivaScope imaging system, either for the diagnostic assessment of lesions suspicious of melanoma or BCC or for the margin delineation of LM prior to surgical treatment.

The costs associated with examination of skin lesions with VivaScope comprise the purchase (capital) cost of the VivaScope imaging system, maintenance costs, the costs of equipment parts and other consumables required for the examination, and the costs of training staff in operating the system and in the assessment and interpretation of the images obtained. They also include costs of staff time required for the examination with VivaScope and subsequent assessment of skin lesions.

The company provided the purchase price of VivaScope 1500 and 3000, as well as the annual maintenance costs. The purchase price and annual maintenance costs of VivaScope 3000 as an add-on device to VivaScope 1500 were stated to be lower than the corresponding costs of VivaScope 3000 as a stand-alone device. For the use of VivaScope in the diagnostic assessment of lesions suspicious of skin cancer, VivaScope 1500 and 3000 have been considered to be used according to indications and suitability (i.e. the economic models assumed that VivaScope 1500 is used for the diagnostic assessment of body skin lesions, whereas VivaScope 3000 is used for the diagnosis of skin lesions on the head or neck). Thus, the lower purchase price and annual maintenance costs for VivaScope 3000 as an add-on device to VivaScope 1500 were utilised in the estimation of costs in all economic analyses that included any of the diagnostic ‘part’ economic models. In contrast, mapping of LMs prior to surgical treatment can be achieved only with the use of VivaScope 3000. Therefore, in the analysis that assessed the cost-effectiveness of VivaScope used exclusively for presurgical margin delineation of LMs, the purchase and annual maintenance costs of VivaScope 3000 as a stand-alone device were used.

The purchase price of VivaScope 1500 and 3000 was annuitised over the expected lifetime of the technology, which was reported by the company to be 10 years. The equivalent annual cost was calculated from the purchase price of the technology and the useful life of the equipment, as advised by the company, using an inflation rate of costs of 3.5%.

The costs of parts refer to the cost of the tissue ring and the cost of a cap for VivaScope 3000. The purchase price of VivaScope includes two tissue rings and two caps. Extra parts need to be purchased only in the event of loss or damage and, therefore, were not considered in the estimation of the total cost of VivaScope.

The consumables required for an examination of a skin lesion with VivaScope include, according to the manufacturer, an adhesive window that is attached to the lesion only for examination with VivaScope 1500; crodamol oil (Croda International Plc, Goole, UK) used as a lubricant; Alcotip sachets (Universal Hospital Supplies, Enfield, UK), which are used for the preparation (disinfection) of the skin; and ultrasound gel. MAVIG GmbH (Munich, Germany), the company that manufactures VivaScope, provided the cost of adhesive windows. For the other consumables, a small cost per lesion examined was assumed, estimated after considering the market prices of the consumables and the fact that only a small portion of each is required per lesion examination.

Training on the use of VivaScope consists of the following (information provided by the company, supplemented by one of the experts providing the training):

• Introductory training This is provided on-site for free with the purchase of VivaScope, lasts approximately 1–2 days and involves mainly technical training, although some basic clinical information is also offered. The purpose of training is to teach technicians and clinicians (consultant dermatologist, consultant dermatological surgeon, technical assistant, pathologist and researcher) how to use the machine and the software correctly, how to identify the anatomical location of the image on the monitor and how to detect the most common and evident structures. Participants are given information on image acquisition, data management, operational precautions, etc. The training course consists of presentations, the review of manuals, the discussion of imaging guidelines and the consideration of appropriate studies of interest.

• Independent study with textbooks This is complementary to the introductory training; VivaScope users are expected to revise two sophisticated imaging textbooks.

• Intensive expert training This is also provided free with the purchase of VivaScope and follows the introductory training and independent study. It is a 3-day course currently offered four times a year at the University of Modena and Reggio Emilia in Italy, but there are plans to expand it to referral centres in Europe, including the UK. Four confocal experts who have been working with the VivaScope for > 10 years in Italy provide the training. They guide participants through the diagnosis of melanocytic lesions, non-melanocytic lesions, inflammatory skin diseases, cosmetic applications and others. It is considered an essential part of the training.

• Online training course Provided free with the purchase of VivaScope, this course consists of 100 cases, with expert evaluation made available after student evaluation. It is considered part of the intensive expert training. The aim of this course is to establish the learning and test the trainee’s skills.

According to clinical expert opinion, after this ‘first-degree’ level of training, which usually lasts 3–5 weeks, trainees are able to recognise features, describe cases and identify diagnoses following algorithms, but they cannot be considered fully trained for routine activity (i.e. they cannot fully achieve the clinical advantages offered by optimal use of VivaScope). Clinicians trained in the use of VivaScope will need to develop their skills further and gain experience and a good level of confidence in interpreting VivaScope images before they achieve the outcomes described in the literature following examination of skin lesions with VivaScope.

At the University of Modena and Reggio Emilia in Italy, this ‘first degree’ of training is followed by a ‘second degree’ of training, consisting of an intense teaching programme, with a duration ranging from 3 to 5 months. This includes a total of approximately 30 hours of teaching (including a short basic course on histopathology), 50 hours of ‘tutored cases’ (case review with an expert and group discussion of the cases) and 100 hours of activities in the skin cancer unit (systematically using confocal microscopy). After this programme, the trainees should achieve a consistent increase in confidence (translated into clinical benefits from VivaScope use that is comparable to literature data), and also a reduction in some initial mistakes in the management of difficult situations (such as pink lesions, undefined papule/nodules, etc.).

According to clinical experts with experience in the use of VivaScope, the overall training required for a clinician to reach a good level of confidence and expertise, is between 4 and 6 months’ time, and approximately 1000–2000 cases evaluated with confocal microscopy in a setting including a sufficient number of melanomas (> 200). This is broadly consistent with the view expressed by Pellacani et al. ,46 according to which ‘a minimum 6 months full-time training, including the evaluation of more than 4000 cases, is required to obtain adequate levels of diagnostic accuracy and confidence’.

Further to the above training, MAVIG GmbH indicated the availability of the following services:

• Online expert tutorial Clinicians may send very difficult confocal cases arising uring the daily clinical practice to a confocal expert for a ‘second opinion’. In this way, clinicians may expand their knowledge and increase their ability to diagnose difficult-to-assess lesions with a high degree of reliability and accuracy. This service is intended as an educational tool and requires a revised VivaNet telemedicine service (Lucid, Inc., Rochester, NY, USA).

• Independent international circle of experts This is a group of expert VivaScope users that offers interdisciplinary discussions in order to establish confocal laser scanning microscopy as the standard in the dermatological diagnosis.

Estimation of training costs and staff time required for examination of skin lesions with VivaScope has been based on the information described above regarding the training courses available and expert advice, according to which a VivaScope facility run by a skin cancer MDT service requires staffing with a band 7 radiographer, who is sufficiently qualified to interpret images, and a well-trained consultant dermatologist or specialist registrar.

The estimation of training costs for the purpose of this evaluation has been based on the information provided by the company regarding the ‘first-level’ training (introductory training and intensive expert training course). No course fees have been considered, as both courses are provided free with the purchase of VivaScope. In terms of staff time, 1.5 days of two radiographers and two dermatologists (for the introductory training) and a further 4 days of two dermatologists (for the 3-day intensive expert training plus travel time to/from Italy) were included in the cost. Moreover, £2000 of travel, hotel and subsistence costs for each dermatologist attending the intensive expert training was included in the estimation of training costs.

It should be noted, however, that the estimate of training costs above does not take into account the substantial further time of ongoing training during routine clinical practice (about 3–5 months) that is required before dermatologists acquire enough confidence and expertise to achieve the full clinical benefits resulting from the use of VivaScope. This means that the conclusions of the economic analysis undertaken to support this report, which has utilised optimal diagnostic accuracy data for VivaScope (as reported in relevant applications), are applicable after dermatologists using VivaScope obtain a good level of expertise (i.e. at about 3–5 months of routine clinical practice following training) in order to achieve the outcomes reported for VivaScope in the literature.

No further training costs for new radiographers and dermatologists using VivaScope in the future were considered, as it has been assumed that the radiographers and dermatologists who were originally trained can subsequently train and pass their experience on to new colleagues expected to use the device on-site and during routine clinical practice.

The total estimated training costs were annuitised over 10 years (equal to the expected lifetime of the device). The equivalent annual cost was calculated using an inflation rate of 3.5%.

In terms of staff time, clinical experts with experience in using VivaScope indicated that examination of skin lesions suspicious of cancer with VivaScope 1500 requires 10 minutes of a radiographer’s time (from the time the patient enters the consultation room until the end of visit, including a radiographer’s time for attaching the adhesive window and obtaining the image) plus 5 minutes of a dermatologist’s time for evaluation of images. Examination of skin lesions suspicious of cancer with VivaScope 3000 requires 10 minutes of a dermatologist’s time (from the time the patient enters the consultation room until the end of the visit, including a dermatologist’s time for obtaining and interpreting the image). In the case of patients with more than one suspected lesion, it was assumed that 50% of a radiographer’s and dermatologist’s time (from the time the patient enters the consultation room until the end of the visit) was fixed, and the remaining 50% was attributed to each lesion examined. Mapping of LMs with VivaScope 3000 prior to surgical treatment requires 30 minutes of a dermatologist’s time.

The unit costs of radiographers and dermatologists were taken from national sources. 65 The unit cost of a hospital band 5 radiographer was £38 per hour in 2014, based on the mean full-time equivalent basic salary for Agenda for Change band 5 for qualified allied health professionals of the July 2013–June 2014 NHS staff earnings estimates. This unit cost included salary (considering also overtime, shift work and geographic allowances) and salary oncosts, capital and other overheads, and qualification costs. The mean annual basic pay of Agenda for Change band 7 qualified allied health professionals is approximately 63% higher than the corresponding pay at Agenda for Change band 565 and, therefore, the unit cost for a band 7 radiographer was estimated to equal £62 per hour (this was estimated by the EAG, as no relevant figures were available in the literature).

The unit cost of a medical consultant was £140 per contract hour in 2014, including, as above, salary and salary oncosts, capital and other overheads, and qualification costs. 65 The unit cost of a specialist registrar was not available (a mean unit cost was available for the registrar group, which comprise a heterogeneous group of registrars, senior registrars, specialist and specialty registrars). The unit cost of an associate specialist was reported to be £124 per hour (under a 40-hour week). For costing purposes, the economic analysis assumed that a consultant dermatologist performs the clinical examination with VivaScope.

The equivalent annual purchase and training and annual maintenance costs of VivaScope were each divided by the annual number of cases (skin lesions) expected to be examined with VivaScope 1500 or VivaScope 3000 for either diagnosis or margin delineation in a skin cancer MDT service in order to distribute these costs across the lesions examined and estimate an annual fixed and training cost per examined lesion. The cost of adhesive windows was omitted from the cost of suspicious skin lesions on the head or neck, as these are examined with VivaScope 3000. As reported in Introduction: overview of methods, it was estimated that approximately 100 suspected melanomas with equivocal dermoscopy findings, 500 lesions suspicious of BCC and 75 diagnosed LMs prior to surgical treatment are eligible for examination with VivaScope by a skin cancer MDT service over 1 year. The percentage of equivocal lesions suspicious of melanoma and of suspected BCC lesions that are on the head or neck was estimated to be 17.9% and 69.4%, respectively.

The economic analysis that assessed the integrated use of VivaScope 1500 and 3000 for the diagnostic assessment of equivocal lesions suspicious of melanoma considered exclusively a number of 100 lesions per year. The analysis that assessed use of VivaScope for diagnostic assessment of lesions suspicious of BCC used a number of 500 lesions per year. The economic analysis considering diagnostic assessment of both types of lesions suspicious of skin cancer with VivaScope assumed an annual number of 600 lesions. The analysis that evaluated the use of VivaScope 3000 in the margin delineation of LMs prior to surgical treatment assumed an annual number of 75 lesions. Finally, the overall analysis of all three uses of VivaScope imaging system that were considered in the economic modelling undertaken for this report utilised an annual number of 675 lesions eligible for examination with VivaScope.

The cost of VivaScope examination per skin lesion examined, by type of skin lesion and analysis considered, is shown in Table 24.

Study population

The study population for this model comprised people with lesions suspicious of melanoma and an equivocal finding on dermoscopy.

The BAD guidelines on management of cutaneous melanoma define populations at greatly increased risk of melanoma (> 10 times that of the general population). 13 These include people with a giant congenital pigmented hairy naevus (such as ≥ 20 cm in diameter or 5% of body surface area), people with a strong family history of melanoma or pancreatic cancer (three or more family members), people with two family members affected with melanoma who also have the atypical mole syndrome, or a history of multiple primary melanomas in an individual or pancreatic cancer. This very high-risk subgroup of patients requires regular monitoring (approximately every 6 months), often over a lifetime, as the risk of some of their skin lesions being malignant or their risk of developing a new melanoma over time is high. In very high-risk patient subgroups with multiple lesions, current practice is selection and excision of a number of lesions based on dermoscopy and clinical judgement, and monitoring of the remaining lesions, as it is not possible to excise all suspicious lesions. If a melanoma is not identified, it will probably be picked up during routine monitoring within 6 months to 1 year. Examination with VivaScope would be beneficial in this subgroup of patients, as it would help identify melanomas among the suspicious lesions so that they are excised earlier rather than later and also would help avoid unnecessary diagnostic biopsies of non-malignant lesions. These very high-risk subpopulations were not considered in the economic model as their management (routine monitoring) differs from that of the ‘average’ population with suspected melanoma; populations at greatly increased risk of melanoma make up a very small proportion of people at risk of melanoma, whose management, nevertheless, can be very resource intensive.

Other categories of moderately increased risk patients (approximately 8–10 times that of the general population) include organ transplant recipients, those with either a previous primary melanoma or a large numbers of moles, some of which may be clinically atypical changing naevi, as well as people with other risk factors for melanoma (e.g. aged ≥ 50 years, prior history of cancer), for whom long-term follow-up is not routinely recommended, were included in the study population of the analysis.

The mean age of the study population, that is, people with suspected melanoma with an equivocal finding on dermoscopy, was assumed to be the same as that of the population receiving a diagnosis of melanoma. Clinical expert advice was that the mean age of people with equivocal lesions suspicious of melanoma does not differ from the mean age of people with lesions suspicious of melanoma in general. Malignant melanoma incidence is related to age, but it has an unusual pattern compared with other types of cancer. In the UK, between 2008 and 2010, an average of 27% of cases were diagnosed in those aged < 50 years, and an average of 45% of cases were diagnosed in those aged ≥ 65 years. 2 Age-specific incidence rates increase steadily from around age 20–24 years, reaching a peak at age ≥ 85 years for both sexes (with the increase being sharper for males from age 55–59 years onwards). 2 The mean age of patients at presentation of melanoma has been reported to be 55 years, although different types of melanoma typically present at different ages. 77 A retrospective study of 1769 people with melanoma who had been referred to a tertiary centre in London from 1999 to 2012 showed that the mean age of patients was 58 years. 78 Using the available information, the age of the study population in the economic model was assumed to be 55 years.

In 2011 there were 13,348 new cases of melanoma in the UK, 6495 of which were in males,2 so that the proportion of men in the population of people with newly diagnosed melanoma was 48.7%. This figure was used in the economic model to reflect the percentage of men in the population with suspected melanoma following an equivocal finding on dermoscopy as well as the population of men with (identified or non-identified) melanoma.

The proportion of confirmed melanomas on the head or neck is approximately 14% in females and 22% in males. 71 These figures were also used to reflect the proportion of suspected and confirmed melanomas with an equivocal finding that are on the head or neck in women and men, respectively.

Each person with suspected melanoma may present with more than one equivocal lesion on dermoscopy, although clinical experts advised that the majority of people present with only one lesion suspicious of melanoma. In studies included in the systematic review of clinical evidence that assessed the diagnostic accuracy of VivaScope in identification of melanomas among equivocal lesions and reported both number of study participants and number of equivocal lesions, the number of suspected melanomas with an equivocal finding per person ranged from 1.0030 to 1.17. 42 However, the number of confirmed melanomas per person reported in the studies was 1, and this was in agreement with clinical expert opinion. For simplicity purposes, the economic model assumed that every person presents with one suspected melanoma with equivocal finding on dermoscopy.

The annual number of equivocal lesions suspicious of melanoma examined at a dermatology MDT service in the UK was estimated to be approximately 100, as reported in Annual number of cases eligible for examination with VivaScope in a dermatology multidisciplinary team clinic in the UK.

Intervention and comparator

The intervention assessed in this model was VivaScope 1500 (for body lesions) and VivaScope 3000 (for lesions on the head or neck) for the diagnostic assessment of skin lesions suspicious of melanoma, following an equivocal finding on dermoscopy. The comparator was routine management of equivocal lesions suspicious of melanoma, comprising excision and biopsy for the majority of the equivocal lesions (highly suspicious lesions) and monitoring for the rest of them (moderately/low suspicious lesions). Monitoring consisted of one outpatient dermatology visit at 3 months, followed by discharge if there was no indication of melanoma.

Model structure

A decision tree followed by a Markov model was constructed to assess the cost-effectiveness of VivaScope in the diagnosis of people with lesions suspicious of melanoma following an equivocal finding on dermoscopy. According to the model structure, which was determined by clinical expert advice and the availability of relevant data, people aged 55 years with dermoscopically equivocal lesions suspicious of melanoma were either examined with VivaScope 1500 or VivaScope 3000 as appropriate (according to the location of the lesion) or received routine management, comprising excision and biopsy of the majority of the suspicious lesions and monitoring of a smaller proportion of less suspicious ones.

The model assumed that confirmed cases of skin cancer are of the same type of cancer as initially suspected (in the case of this model, melanoma), although occasionally skin cancers identified may be of a different type from that initially estimated by the clinician via dermoscopy.

People whose lesions were examined with VivaScope received the results of the examination immediately. Lesions found to be positive on VivaScope examination underwent excision and biopsy. The results of the biopsy were received 2 weeks after excision. Lesions found to be TP (i.e. confirmed on biopsy to be melanoma) were further treated as melanoma according to stage, as recommended by national guidelines. Those that were FP (i.e. biopsy showed it was not a melanoma) were assumed to be a benign tumour that did not require treatment and patients were discharged after the (unnecessary) excision and biopsy. People whose lesions were found to be negative on VivaScope examination were discharged and advised to visit their GP if they noticed changes in their skin lesion. If the lesions were TN (i.e. not melanoma), then they were assumed to be a benign tumour that did not require treatment. If the lesions were FN (i.e. an unidentified melanoma), it was assumed that the patient would return to the service at a later time once the melanoma had potentially progressed to a more advanced stage.

People who received routine management, whose lesions were excised and biopsied, received the results of biopsy 2 weeks after the excision. Those who had a positive result (i.e. their lesion was confirmed be a melanoma) were treated for their melanoma according to its stage, as recommended by national guidelines. Those who had a negative result (i.e. biopsy showed that their lesion was not a melanoma) were assumed to have a benign tumour that did not require treatment and were discharged after the (unnecessary) excision and biopsy. People under routine management who were selected for monitoring attended an outpatient dermatology follow-up appointment at 3 months for re-evaluation of their lesion. If the lesion was found to be suspicious of melanoma, it underwent excision and biopsy, which was followed by either further appropriate treatment, if biopsy confirmed the presence of malignancy, or discharge, if the result of biopsy was negative. If at the follow-up appointment the lesion was found not to be suspicious, patients were discharged and advised to visit their GP if they noticed changes in their skin lesion. If the lesion was not malignant, patients were assumed not to require further treatment. If the lesion was malignant but was not identified at the follow-up meeting, it was assumed that the patient would return to the service at a later time once the melanoma had potentially progressed to a more advanced stage. However, if a malignant lesion was identified at the 3-month follow-up meeting, it was assumed not to have progressed to a more advanced stage.

All people undergoing excision and biopsy of their lesion experienced distress because of the procedure; they also experienced anxiety while waiting for the results of biopsy, whether or not they had been examined with VivaScope prior to excision. People with a FP result after VivaScope examination experienced anxiety thinking that they have melanoma, until the results of biopsy were available.

Following the outcome of the diagnostic assessment, people entered a Markov model and followed one of the following pathways:

1. Patients with a confirmed melanoma (i.e. those with a TP result after VivaScope examination and subsequent excision and biopsy, as well as those who, under routine management, had a positive result after excision and biopsy, either immediately or following monitoring) entered a Markov chain of ‘identified melanomas’. All melanomas were assumed to be in situ or stage I (Ia or Ib) at identification, as clinical experts advised that an equivocal finding on dermoscopy suggests early stages of melanoma. All identified melanomas were treated in accordance with national guidelines and were assumed not to progress to a more advanced stage. Patients with an identified melanoma had a reduction in their HRQoL. A proportion of those who had a melanoma on their head or neck experienced an additional permanent reduction in their HRQoL because of the scarring following excision and biopsy. Patients with an identified melanoma stage Ib were at increased risk of mortality, because of their melanoma, for the first 10 years following identification of their melanoma. After the period of 10 years, the risk of mortality of people with identified melanomas returned to that of the general population of the same age. People dying because of their melanoma were assumed to become terminally ill in the year in which they died.

2. Patients with a missed melanoma (i.e. those with a FN result after VivaScope examination, as well as those who, under routine management, were selected for monitoring and were not identified) entered a Markov chain of ‘non-identified melanomas’. All melanomas were assumed to be in situ or stage I (Ia or Ib) at the point of examination; however, they could progress to more advanced stages over time. Every year patients could remain in their undiagnosed status with their melanoma remaining at the same stage or progressing to the next stage (without incurring any costs for its management), or could return to the dermatology service because of changes in their lesion and be diagnosed and treated, or die because of their cancer. Clinical experts advised that any unidentified melanomas would be recognised by the time they reached stage II (IIa, IIb or IIc), and within 5 years at maximum after the initial examination that resulted in the equivocal dermoscopic finding. People with an unidentified melanoma had a HRQoL equal to that of the general population of the same age, until their melanoma was identified, in which case they experienced a reduction in their HRQoL. A proportion of those who had an identified melanoma on their head or neck experienced an additional permanent reduction in their HRQoL because of the scarring following excision and biopsy. Unidentified melanomas did not incur any costs; identified melanomas were treated in accordance with national guidelines and were assumed not to progress to a more advanced stage. People with an unidentified or identified melanoma at stage Ib or II were at increased risk of mortality, because of their melanoma, from the start of the model and for the first 10 years after identification of their melanoma. After that period, their risk of mortality became equal to that of the general population of the same age. People dying because of their melanoma were assumed to become terminally ill in the year in which they died. People with newly identified melanomas entered tunnel states over a period of 10 years, so that the time-dependent risk of mortality over that period could be applied. In this aspect, the economic model was not a Markovian one under a strict definition, given that tunnel states allowed the model to keep memory of the time period people spent with melanoma, once this was identified.

3. People without a melanoma (i.e. those with a FP or TN result after VivaScope examination, as well as those who, under routine management, had a negative result after excision and biopsy, either immediately or following monitoring) entered a Markov chain of ‘no melanomas’. A proportion of people with a benign lesion on their head or neck, who had undergone unnecessary excision and biopsy, experienced a permanent reduction in their HRQoL because of the resulting scarring. Otherwise, the HRQoL in this Markov chain and the mortality risk were equal to that of the general population of the same age.

The care pathways described above were adapted from Wilson et al. ,48 who developed an economic model to assess the cost-effectiveness of a device aiming at the diagnostic assessment of pigmented skin lesions in primary care in the UK. The pathways designed for the model developed for this report were finalised following clinical expert advice.

Management of identified melanomas comprised surgical excision with a wider and deeper margin for all melanomas, SLNB for 50% of melanomas of stage Ib and all stage II melanomas, and follow-up visits. Patients dying because of their cancer incurred terminal disease costs in the year in which they died. These included costs of radiological examination, costs of metastatic disease (costs of chemotherapy including costs of adverse events), inpatient care and outpatient attendances, as well as costs of terminal and palliative care.

The time horizon of the economic model was over a lifetime (up to 100 years of age). The cycle length of the Markov model was 1 year and half-cycle correction was applied.

Figure 6 shows a schematic diagram of the VivaScope diagnostics model on suspected melanoma following an equivocal finding on dermoscopy.

FIGURE 6.

Schematic diagram of the VivaScope diagnostics model on suspected melanoma following an equivocal finding on dermoscopy. (a) Decision tree component; (b) progression of people with melanomas following identification (VivaScope TP or identified at biopsy or during monitoring); (c) progression of people with melanomas following initial non-identification (VivaScope FN or not identified during monitoring); and (d) progression of people without melanoma (VivaScope FP or TN, or identified as negative following biopsy or monitoring).

Clinical input parameters

Utility values

People in the model experienced utility (or disutility) associated with one or more of the following:

• disutility because of the excision and biopsy of a lesion suspected of melanoma that caused distress as well as anxiety while waiting for the results

• disutility because of the permanent scarring following surgical excision of a lesion on the head or neck

• health state-related utility, which was associated with the stage of melanoma (in people with melanoma) or with the average utility of the general population (in people without a melanoma).

As reported in Studies reporting utility data, the systematic literature review identified four studies reporting utility data relating to melanoma health states. 53–55,58 Beusterien et al. 54 reported utility data associated with partial response to treatment, stable or progressive disease, best supportive care and toxicity from chemotherapy. The health state descriptions were vignette based. Utility values were elicited from members of the general population in Australia and the UK using SG. The utility values reported by Beusterien et al. 54 referred to health states that did not directly correspond to melanoma stages and, therefore, were unsuitable for use in the economic model.

The remaining three studies53,55,58 reported utility values associated with melanoma stages. None of the studies was conducted in the UK. Two of the studies53,58 used EQ-5D for the description of HRQoL experienced by patients with melanoma. However, none of them used the UK EQ-5D tariff62 for the valuation of health states, as recommended by NICE. Askew et al. 53 used the US EQ-5D tariff, which was developed using TTO, whereas Tromme et al. 58 used the Belgian EQ-5D tariff, which was developed using the VAS – a valuation method that is not choice based and thus is not among NICE-preferred valuation methods. King et al. 55 reported melanoma-related utility values elicited from patients with melanoma in the USA; health state descriptions were based on vignettes. A comparison of the utility values reported in these three studies revealed inconsistencies in the available data. For example, the utility values reported by Askew et al. 53 for melanoma stages III and IV were considerably higher than those reported by King et al. 55 and Tromme et al. ;58 the utility values reported by Tromme et al. 58 for melanoma early stages I and II were substantially lower than the utility values reported for corresponding stages in Askew et al. 53 and King et al. 55 These discrepancies are potentially attributable to differences in measurement and valuation across the three studies. Quite importantly, the utility values reported for all melanoma stages (I–IV) in Askew et al. 53 and for stages I and II in King et al. 55 were considerably higher than reported mean utility values for the UK general population aged 55 years (which was the age of the study population at the start of the model): in Askew et al. ,53 the utility values of melanoma stages I–IV ranged from 0.91 in stages I and II to 0.86 in stage IV. King et al. 55 reported utility values of 0.93 and 0.92 for melanoma stages I and II, respectively. In contrast, Kind et al. ,85 who analysed EQ-5D data obtained from 3395 participants in the measurement and valuation of health survey conducted in the UK in 1993, reported a mean EQ-5D utility value for people in the UK aged 55–64 years of 0.78 for men and 0.81 for women. More recently, Sullivan et al. 86 produced a catalogue of EQ-5D utilities for the UK population by applying the UK EQ-5D tariff62 to EQ-5D descriptive questionnaire responses obtained from participants in the US-based Medical Expenditure Panel Survey. The mean utility value for people aged 50–59 years was 0.798. Consequently, the utility data reported in Askew et al. 53 and King et al. 55 appeared to lack face validity compared with UK population norms and could not be used in the economic model in their ‘raw’ form, as this would result in patients with melanoma having a higher utility than people without melanoma (who are expected to have the utility of the general population of same sex and age).

The other utility study under consideration was the one conducted by Tromme et al. 58 The study reported utility values associated with melanoma stages 0 (in situ)/Ia, Ib/II, III and IV, subdivided into treatment and remission phases. The reported utility values appeared to be sound when compared with mean utility values of the UK general population, as the utility values of treatment phase were always lower than the utility of the UK general population aged 55 years (mean age of patients at presentation of melanoma) and utility values of remission phase were lower than (stages III and IV), or comparable with (stages 0–II), the utility of the UK general population aged 55 years. Utility values reported in Tromme et al. 58 were estimated from EQ-5D responses using the Belgian EQ-5D tariff, which has been developed following a valuation survey of 2754 Flemish adults from the general public in Belgium using the VAS. 61 The Flemish EQ-5D tariff has shown good correlation with the UK EQ-5D tariff that was derived using the VAS,87 with a correlation coefficient of 0.979. 61 Owing to the lack of melanoma utility data more directly relevant to the UK population, melanoma-related utility values from Tromme et al. 58 were selected for use in the economic model.

The utility values obtained from Tromme et al. 58 were adjusted for age in the economic analysis, using a coefficient of –0.00029 per year that was reported by Sullivan et al. ;86 this study involved multiple regression analyses using ordinary least squares, the Tobit model and censored least absolute deviations regression methods and reported regression coefficients for a number of clinical conditions and demographic characteristics of the study population, including age.

Table 26 shows the patient characteristics and mean utility values by melanoma stage reported in Tromme et al. ,58 as well as the resulting utility values for patients with melanoma aged 55 years, after adjusting for age, that were used in the economic model.

The utility values of stages 0/Ia and Ib/II in remission reported by Tromme et al. 58 were very close to (only slightly higher than) the utility values of the UK general population aged 55 years reported by Kind et al. 85 and Sullivan et al. 86 Tromme et al. 58 reported that treatment duration in stages 0/Ia and Ib/II was 1 and 2 months, respectively, according to expert opinion. Using the treatment and remission utility data and the treatment duration for stages 0/Ia and Ib/II, it was estimated that the reduction in utility over the year within the melanoma was treated was –0.0100 for stage 0/Ia and –0.0371 for stage Ib/II. Tromme et al. 58 stated that patients with stage 0 and Ia melanoma were pooled together because they had very similar management in terms of surgical treatment and follow-up, whereas patients with stage Ib and II melanoma were pooled because they had all undergone SLNB that had not been followed by elective node dissection and because of evidence that surgical resection margins did not appear to influence HRQoL. Therefore, the values of –0.0100 for stage 0/Ia and –0.0371 for stage Ib/II were considered to express the disutility associated with surgical management of early stage melanoma which involved (–0.0371) or did not involve (–0.0100) SLNB in patients with melanoma aged 55 years. These values were applied as one-off disutilities in the economic model (i.e. they were applied once, at the time of treatment of melanomas, without time adjustment), after adjusting for age by applying the age coefficient of –0.00029 reported by Sullivan et al. ,86 for every year above 55 years of age. However, as only 50% of patients with stage Ib melanoma were assumed to undergo SLNB in the economic model, the value of –0.0100 (adjusted for age) was applied to all patients with identified melanoma of stage 0/Ia and 50% of patients with melanoma of stage Ib, and the value of –0.0371 (adjusted for age) was applied to 50% of patients with stage Ib melanoma and all patients with stage II melanoma. Apart from that disutility, which was attributed to surgical management and was applied as a one-off disutility, patients with identified melanoma in stages 0–II were assumed to have the average utility of the UK general population of the same age, which was derived from Sullivan et al. 86 as the utility of stages 0–II in remission reported in Tromme et al. 58 was very close to (in fact it was higher than) the utility of the UK general population reported in Sullivan et al. 86 This assumption is broadly consistent with the results of a German study, according to which the HRQoL in patients with melanoma, without recurrence within 2 years after initial therapy, was comparable to the HRQoL of the general population. 60

Patients with unidentified melanoma and people without melanoma were also assumed to have the average utility of the general UK population of the same age, as reported in Sullivan et al. 86

Table 27 presents the mean utility of the UK general population aged ≥ 55 years, as reported in Sullivan et al. 86 and applied in the economic model, as well as the characteristics of the US population providing responses to EQ-5D that were analysed by Sullivan et al. 86 in order to produce the catalogue of UK utility values. The mean EQ-5D utility for people aged 80–89 years was applied to all people aged ≥ 80 years in the economic model.

People with metastatic melanoma disease/terminal illness (i.e. people dying because of their melanoma) were assumed to have the utility of melanoma stage IV in treatment reported in Tromme et al. ,58 which was adjusted for age using the age coefficient of –0.00029 reported by Sullivan et al. ,86 for every year above 55 years of age.

People undergoing surgical excision and biopsy of their lesion were assumed to experience disutility because of distress as well as anxiety while waiting for the results of the biopsy. The distress because of the excision and biopsy experienced by people whose suspected lesion was melanoma was assumed to have been incorporated in the disutility associated with the surgical management of melanoma. The distress experienced by people whose suspected lesion was not melanoma (FN), who therefore did not proceed to surgical excision with wider margins, was expressed by a one-off disutility of –0.002, which was also used to express the disutility of a diagnostic biopsy for suspected BCC in the relevant economic model. As described later in Utility values, the disutility experienced because of surgical treatment of BCC was derived from Seidler et al. ,57 who reported a disutility of –0.004 associated with an excision procedure because of facial non-melanoma skin cancer. The economic model on lesions suspicious of BCC assumed that a diagnostic biopsy created a disutility of –0.002 to the person, as it was expected to be a less invasive procedure than surgical treatment of BCC (excision or Mohs surgery).

In addition to the distress directly associated with excision and biopsy, people undergoing excision and biopsy for their suspected melanoma lesion were considered to experience a reduction in their HRQoL because of anxiety while waiting for the results of biopsy. The methodology used to estimate the disutility associated with anxiety while waiting for results of biopsy was adopted from a model-based economic evaluation of intraoperative tests for detecting sentinel lymph node metastases in breast cancer88 that was undertaken to inform relevant NICE diagnostics guidance. In that economic model, patients who underwent histopathology experienced some level of disutility because of the associated anxiety of waiting for test results; this disutility was imputed by using the EQ-5D health state valuation equation for the UK reported by Dolan,62 which allows an estimation of a person’s utility based on their responses to the EQ-5D classification system. The system has five dimensions (mobility, self-care, ability to perform usual activities, pain/discomfort and anxiety/depression) and in the version used by Dolan each dimension had three levels of response (no problems, moderate problems and severe problems). Huxley et al. 88 used only the utility decrement because of anxiety/depression, which was expressed by the following equation:

where α = 0.081 is the constant applied to any level of disutility in any of the five EQ-5D dimensions, AD = 0.071 (for each level of disutility associated with anxiety or depression), A2 = 0.094 (for severe anxiety/depression) and N3 = 0.269 (when any of the five dimensions of EQ-5D is severe).

Huxley et al. ,88 as well as the economic model on the diagnostic assessment of equivocal lesions suspicious of melanoma with VivaScope, assumed that people waiting for histopathology results already had a utility of < 1 (so the α value was not applied at the estimation of the utility decrement because of anxiety/depression), that they moved from a state of no anxiety/depression to severe anxiety/depression, and that this anxiety/depression was the only dimension of the EQ-5D they had that was severe. These assumptions resulted in a disutility of –0.505 (i.e. –0.236 – 0.269 = –0.505).

This disutility of –0.505 was applied for only 2 weeks in the model, as clinical experts advised that biopsy results for suspected melanoma are available 2 weeks after excision and biopsy. This gave a 2-week disutility of –0.019 attributed to anxiety while waiting for the results of biopsy. This disutility was applied in every person waiting for results of biopsy, including people who had already undergone examination with VivaScope, people undergoing routine management of equivocal lesions suspicious of melanoma who had their lesions excised immediately or after 3-month monitoring, and people with missed melanomas (FN) who had them excised at a later stage following late identification.

A number of people in the model experienced permanent disutility because of scars on their head or neck from the excision of suspected melanomas. In the economic model it was assumed that 15% of people undergoing excision of their suspected melanoma lesion on their head or neck would experience permanent disutility because of their scar over their lifetime. Seidler et al. 57 reported a disutility of –0.016 for simple repairs/scars (granulation and primary closure) and a disutility of –0.026 for complex repairs/scars (local flap and graft) experienced by people with facial non-melanoma skin cancer. Owing to a lack of more relevant data, data from this study were used to express permanent disutility experienced by people with suspected melanomas on their head or neck because of scars from excision. Clinical expert advice was that all initial excisions of suspected melanomas are undertaken with simple repairs/scars; wider surgical excisions of confirmed melanomas make up 90% of simple and 10% of complex repairs/scars. Based on these estimates and the disutility data reported in Seidler et al. ,57 the permanent disutility from scarring following initial excision and biopsy (people with lesions that were not melanomas) and wider surgical excision (people with melanomas) was estimated to be –0.016 and –0.017, respectively. These disutilities were applied only to people with permanent reduction in their HRQoL because of scarring on their head or neck over their lifetime.

Table 28 provides all utility data applied in the diagnostic economic model on equivocal lesions suspicious of melanoma.

Costs

Costs considered in this economic model included the cost of diagnostic assessment of a suspected melanoma with VivaScope following an equivocal finding on dermoscopy, the cost of routine management (cost of excision or monitoring of suspected melanomas), the management cost of confirmed melanomas (TPs) following diagnostic assessment, the cost of missed melanomas (FNs) that were identified at a later time and costs associated with metastatic melanoma and terminal illness.

As reported in Table 24, the cost per suspected melanoma with an equivocal finding on dermoscopy examined with VivaScope was estimated to be £254 if VivaScope is exclusively used for the diagnostic assessment of suspected melanomas with an equivocal finding on dermoscopy; £63 if VivaScope is used only for diagnostic assessment of suspected melanomas giving an equivocal finding on dermoscopy and suspected BCC lesions with a positive dermoscopic finding; and £59 if the device is used not only for the diagnostic assessment of suspected melanomas and BCCs, but also for the mapping of LMs prior to surgical treatment.

The costs of all other procedures and treatments included in the model, with the exception of the cost associated with terminal illness, were taken from either the NHS Reference Costs 2013 to 201412 for 2014 or the Unit Costs of Health and Social Care 2014. 65 Clinical experts advised on the appropriate NHS service, and procedure codes and unit costs corresponding to relevant health-care resource use considered in the model.

The unit cost of excision and biopsy of a lesion suspicious of melanoma was estimated to be £151, corresponding to the national unit cost of outpatient intermediate skin procedures conducted in a dermatology service for people aged ≥ 13 years (service code 330, currency code JC42A). 12

The unit cost of monitoring was £93 and corresponded to an outpatient, face-to-face, consultant-led dermatology follow-up attendance (service code 330, currency code WF01A). 12

The cost of management of melanomas after identification and confirmation with excision and biopsy (i.e. both melanomas identified at initial diagnostic assessment and melanomas missed and identified at a later time) comprised:

• The cost of surgical excision with a wider and deeper margin for all melanomas (in situ, stage I and stage II), which was £943, corresponding to the national unit cost of an intermediate skin procedure treated as a day case for people aged ≥ 13 years (currency code JC42A). 12 Clinical experts advised that if skin grafts or flaps are required for the excision, the procedure becomes more complex and costly; however, the associated additional cost was not considered because of lack of relevant data.

• The cost of SLNB for 50% of melanomas of stage Ib and all stage II melanomas. The unit cost of such a procedure was estimated to be £1033, corresponding to a day case procedure on the lymphatic system. 12 Clinical experts advised that this procedure is routinely carried out together with the wider excision and, therefore, it might be reasonable not to apply its unit cost as a separate cost component in the model; nevertheless, other experts advised that it can be a complex procedure, especially when performed in complicated nodal sites, for example in the groin or head and neck. Consequently, it was decided to apply the unit cost of £1033 as an extra cost in patients undergoing SLNB alongside the wide surgical excision of their melanoma.

• The cost of follow-up visits: these comprised, according to the BAD guidelines:13

  • a single follow-up visit for patients with in situ melanomas, after complete excision, to explain the diagnosis, check the whole skin for further primary melanomas and to teach self-examination for a new primary melanoma

  • four 3-monthly visits in the first year after the excision of the melanoma for patients with stage Ia melanoma

  • 3-monthly visits for 3 years and then 6-monthly visits to 5 years after the excision of the melanoma for patients with stage Ib or II melanoma.

The unit cost of a follow-up visit was £93 and corresponded to an outpatient, face-to-face, consultant-led dermatology attendance (service code 330, currency code WF01A). 12 It should be noted, however, that clinical experts advised that, in some hospitals, the follow-up of patients with melanoma is nurse led rather than consultant led.

The health-care resource use and associated cost of management of melanomas following excision and biopsy that was utilised in the economic model is presented in Table 29.

The cost of people with unidentified melanomas was assumed to be zero, unless patients died because of their melanoma (in which case they experienced terminal disease before they died and incurred the corresponding costs) or until their melanoma was identified. Costs of identification included a GP visit at a cost of £67,65 an outpatient, face-to-face, consultant-led first attendance at a dermatology clinic for the reassessment of the skin lesion costing £109 (service code 330, currency code WF01B),12 and excision and biopsy for confirmation of the malignancy at a cost of £151.

The cost of terminal illness in the year in which patients died because of their melanoma (cost of management of metastatic disease and terminal care) was based on data reported in the NICE single technology appraisal of ipilimumab for previously untreated advanced (unresectable or metastatic) melanoma (NICE TA319). 89 Based on clinical expert advice, patients with metastatic melanoma and terminal disease in the model were assumed to be treated with either ipilimumab (50%), dacarbazine (15%) or vemurafenib (35%), with the proportions of patients on each drug being based on an economic analysis assessing the cost-effectiveness of adding routine imaging of asymptomatic patients to current standard follow-up in patients with stage III melanoma that was undertaken for the NICE guideline update on melanoma. 14 The drug acquisition costs of ipilimumab and vemurafenib to the NHS are subject to a Patient Access Scheme discount and, therefore, are not known; consequently, it was not possible to estimate the actual costs of chemotherapy to the NHS. The company submission for the NICE TA31989 reported the estimated total metastatic disease and terminal care costs associated with each of the three drugs over a lifetime, as well as the average number of life-years per person for each drug, so it was possible to estimate an average annual cost associated with each drug, although it is acknowledged that costs of chemotherapy and terminal illness are unlikely to be evenly spread across life-years. However, as total lifetime was not long (it did not exceed 3.5 years with any of the three drugs), the estimated mean annual cost was considered a reasonable approximation of metastatic disease/terminal illness cost over the last year of life of patients dying because of their melanoma in the economic model. The single technology appraisal cost figures were derived from a scenario included in the EAG report that considered drugs only as first-line treatments followed by best supportive care and palliative care. 90 These costs included drug acquisition costs, costs of adverse events, costs of radiological examination, inpatient care and outpatient attendances, as well as costs of terminal and palliative care. The metastatic melanoma/terminal disease cost estimated using these data was £16,139, as shown in Table 30.

All other health-care and PSS costs incurred by the study population in the model, including the costs incurred by people with a benign lesion (i.e. people with TN or FP results in diagnostic assessment), were estimated to be equal between the two arms of the model and were thus omitted from the analysis.

Study population

The study population for this model comprised people with suspected BCC lesions with a positive or equivocal result on dermoscopy. The aim of examination of the suspected BCC lesions with VivaScope was to make or confirm diagnosis, as an alternative to diagnostic biopsy.

According to NICE guidance,91 patients with low-risk BCC lesions may be identified and managed by GPs in community care settings. However, clinical experts expressed the opinion that this is not routine practice, and, in reality, GPs manage < 10% of low-risk BCCs; therefore, following clinical expert advice, the economic model assumed that all patients with suspected BCC lesions are referred to (and managed by) specialist dermatologist centres.

The mean age of the study population, that is, people referred to a dermatology department with suspected BCC, was assumed to be the same as the mean age of people at diagnosis of BCC. BCC is more common in older people; people aged > 75 years are about five times more likely to have a BCC than those aged between 50 and 55 years. 2 According to a study that analysed trends in the demographic, clinical and socioeconomic profile of > 50,000 patients with non-melanoma skin cancer registered between 1994 and 2011 by the Irish National Cancer Registry, the median age at diagnosis of BCC was 68 years for both men and women. 74 Another study that analysed data on all cases of BCC diagnosed at a single centre of dermatopathology during 1967–96 in Strasbourg, France, reported that the mean age of people at diagnosis of BCC was 65 years. 73 Data on the mean age of patients with suspected or diagnosed BCC in the UK were not possible to identify, so the mean age of the study population (people with suspected BCC) was estimated to be 63 years, based on the available data and after considering the fact that the age-specific incidence rate for BCCs has been increasing in both sexes for all age groups over the years, with the largest overall increase in BCC incidence rates being observed in the youngest age groups. 74

Non-melanoma skin cancers are more common in males than females in the UK, with a ratio of 13 : 10 (which translates to a proportion of 56.5% males in the total population), although the sex difference is wider for SCC than for BCC. 2 Data from the Irish National Cancer Registry indicated that the proportion of men among patients with BCC between 1994 and 2011 was 52.8%. 74 This figure of 52.8%, which is overall consistent with relevant UK information on non-melanoma skin cancer, was used in the economic model because of a lack of relevant UK data on the male-to-female ratio specific to BCC. This figure was used to represent the percentage of men in the population with suspected (rather than confirmed) BCC following a positive or equivocal finding on dermoscopy.

The proportion of suspected and also confirmed BCC lesions on the head or neck in the model was 69.4%,76 based on information reported in Annual number of cases eligible for examination with VivaScope in a dermatology multidisciplinary team clinic in the UK.

Each person with suspected BCC may present in one visit with more than one lesion that has been found positive or equivocal for BCC on dermoscopy. Clinical experts advised that for non-melanoma skin cancer there is a 50% chance of a second non-melanoma skin cancer in a 5-year period, whereas incidental second tumours may be potentially present in about 10% of patients with BCC. The economic model assumed, for simplicity, that the number of suspected BCC lesions with a positive or equivocal dermoscopic finding per person is equal to the number of confirmed BCC lesions per person; the latter was estimated to be 1.09, using audit data from Teoh et al. ,75 who reported 926 confirmed BCC lesions in 849 patients in a retrospective single-centre audit of all BCC excisions performed in 2008. The figure of 1.09 lesions per person is consistent with clinical expert advice. The study that provided the data on the diagnostic accuracy of VivaScope on suspected BCC lesions for the economic analysis reported a mean number of 1.29 lesions suspicious of BCC per study participant (92 lesions in 72 patients), and a mean number of 1.40 confirmed BCC lesions per person with BCC (45 BCCs in 32 patients). 43 The higher figure of 1.40 BCC lesions per person (suspected or confirmed) was tested in sensitivity analysis. It should be noted that, for purposes of simplicity in the model design, it was assumed that all lesions in one person are either malignant (BCC) or not and follow the same pathway [i.e. received the same result on examination with VivaScope, the same (necessary or unnecessary) treatment], and have the same potential impact on HRQoL.

The annual number of lesions suspicious of BCC with a positive or equivocal dermoscopic finding examined at a dermatology MDT service in the UK was estimated to be approximately 500, as reported in Annual number of cases eligible for examination with VivaScope in a dermatology multidisciplinary team clinic in the UK.

Intervention and comparator

The intervention assessed in this model was VivaScope 1500 (for body lesions suspicious of BCC) and VivaScope 3000 (for suspected BCC lesions on the head or neck) for the diagnostic assessment of skin lesions suspicious of BCC. The comparator was diagnostic biopsy, which was considered to reflect routine practice following a positive or equivocal dermoscopic finding for BCC.

Model structure

A decision tree followed by a Markov model was constructed to assess the cost-effectiveness of VivaScope in the diagnosis of people with lesions suspicious of BCC that had a positive or equivocal finding on dermoscopy. According to the model structure, which was determined by clinical expert advice and availability of relevant data, people aged 63 years with lesions suspicious of BCC following a positive or equivocal finding on dermoscopy were either examined with VivaScope 1500 or VivaScope 3000 as appropriate (according to the location of the lesion) or had a diagnostic biopsy for confirmation of BCC. The model assumed that confirmed cases of skin cancer are of the same type of cancer as initially suspected (in the case of this model, BCC), although occasionally skin cancers identified may be of a different type of that initially estimated by the clinician at dermoscopy.

People whose lesions were examined with VivaScope received the results of the examination immediately. Lesions found positive by VivaScope examination were treated for BCC in accordance with national guidelines; treatment was applied to both TP and FP lesions. Lesions found negative by VivaScope examination underwent diagnostic biopsy (because of the dermoscopic outcome that was suggestive of malignancy), and subsequently received treatment if BCC was confirmed (diagnostic biopsy was considered to be the gold standard for the diagnosis of BCC). The results of diagnostic biopsy were available 6 weeks after the biopsy, in accordance with routine clinical practice in the UK. If the results of biopsy were negative, patients were discharged. It is noted that, under this pathway, no BCC lesions remained undiagnosed, as any FN lesions following VivaScope examination would move on to receive a diagnostic biopsy and would eventually be identified. On the other hand, FP lesions following VivaScope examination received unnecessary treatment.

Lesions assessed with diagnostic biopsy following a positive or equivocal finding on dermoscopy received treatment if BCC was confirmed; otherwise, patients were discharged. The results of diagnostic biopsy were received 6 weeks after the biopsy. All people in this arm of the model received treatment in accordance with their true BCC status and, therefore, none of them received unnecessary treatment.

Treatment of BCC lesions in the model comprised a mixture of surgical and non-surgical therapies, in accordance with published guidelines. 92,93 Surgical therapies included surgical excision and Mohs surgery. Non-surgical treatments included photodynamic therapy, radiotherapy and topical treatment with imiquimod or fluorouracil. Other, overall less common treatments for BCC, such as curettage and cautery, cryotherapy and chemotherapy, were not considered in the economic model. However, it is acknowledged that curettage and cautery, as well as cryotherapy, are commonly used treatments for low-risk BCCs, especially superficial ones. In any case, It should be noted that, according to clinical expert advice, there seems to be variation in clinical practice, with some of the therapies being offered more or less routinely at different dermatology centres across the country.

All people undergoing diagnostic biopsy experienced distress because of the biopsy and anxiety while waiting for the results. All people receiving surgical treatment and those treated unnecessarily with any kind of treatment (surgical or non-surgical) experienced distress because of the treatment. Moreover, a proportion of people undergoing diagnostic biopsy or surgical treatment for skin lesions on the head or neck were assumed to experience a permanent reduction in their HRQoL because of the resulting scarring.

People experiencing a permanent reduction in their HRQoL because of scarring entered a very simple Markov model, consisting of only the states of being alive (with permanent disutility attributable to scarring) and dead, in order to estimate the total disutility because of scarring experienced over a lifetime. Apart from this permanent disutility experienced by a proportion of people in each arm of the model, the choice of diagnostic strategy (i.e. either examination with VivaScope followed by diagnostic biopsy for lesions found negative for BCC or diagnostic biopsy of all suspected BCC lesions) did not have any other impact on costs or outcomes beyond the end of treatment. This is because, in both arms of the model, no BCC remained undiagnosed and, therefore, untreated. Consequently, there was no difference in tumour expansion, recurrence or mortality between the two arms of the model. For this reason, tumour expansion or future recurrence (and associated costs and impact on HRQoL) were not considered in the Markov part of the model. Thus, all future costs and outcomes, with the exception of permanent disutility because of scarring, experienced by a proportion of people, were estimated to be the same in both arms of the model and were therefore omitted from the model. The cycle length of the Markov model was 1 year and half-cycle correction was applied.

The time horizon of the economic model was over a lifetime (up to 100 years of age).

A schematic diagram of the VivaScope diagnostics model on suspected BCC following a positive or equivocal finding on dermoscopy is shown in Figure 7.

FIGURE 7.

Schematic diagram of the VivaScope diagnostics model on suspected BCC following a positive or equivocal finding on dermoscopy. (a) Decision tree component; and (b) the Markov model component.

Clinical input parameters

Utility values

Patients in this model experienced a reduction in their HRQoL for one of the following reasons:

• diagnostic biopsy that caused distress as well as anxiety while waiting for the results

• surgical treatment (all people undergoing surgical excision or Mohs surgery in the model) and unnecessary non-surgical treatment (people with FP lesions)

• permanent scarring following surgical treatment of a lesion on the head or neck.

As reported in Systematic literature review of existing economic evidence, Results, Seidler et al. 57 estimated a disutility of –0.004 associated with an excision procedure because of facial non-melanoma skin cancer using traditional surgical excision or Mohs surgery. They also reported a disutility of –0.016 for simple repairs/scars (granulation and primary closure) and a disutility of –0.026 for complex repairs/scars (local flap and graft).

The study had many limitations and did not meet NICE criteria for use of utility data. Utility values were elicited from five healthy individuals in the USA, who used TTO to value two scenarios relating to surgical excision or Mohs surgery of facial non-melanoma skin cancer. Owing to a lack of better-quality data, the utility values reported in this study were utilised in the economic model. The value of –0.004 was used to reflect the decrement in HRQoL (utility) experienced because of surgical treatment (either surgical excision or Mohs surgery). Owing to the lack of any relevant data on the disutility caused by unnecessary treatment received by people with lesions with a FP result for BCC following examination with VivaScope, it was assumed that the one-off disutility of –0.004 reported in Seidler et al. 57 for surgical treatment applied to any (surgical or non-surgical) unnecessary treatment as well. It was assumed that a diagnostic biopsy created a disutility of –0.002 to the person, as it is expected to be a less invasive procedure than surgical excision or Mohs surgery. The disutility attributable to diagnostic biopsy and the disutility attributable to surgical/unnecessary treatment were applied as one-off disutilities (i.e. they were applied once, at the time of the respective procedure, without time adjustment). These disutilities were assumed to be additive, that is a lesion receiving a diagnostic biopsy followed by surgical treatment created a disutility for the patient of –0.002 + (–0.004) = –0.006 in the year in which it was biopsied and excised. Decrements in utility as a result of diagnostic biopsy or surgical/unnecessary treatment were applied separately to each lesion, so that a person with more than one lesion was assumed to experience a ‘cumulative’ disutility because of procedures experienced on each of their lesions.

In addition to the distress directly associated with diagnostic biopsy, people undergoing a diagnostic biopsy were considered to experience a reduction in their HRQoL because of anxiety while waiting for the results of biopsy. Owing to the lack of any relevant utility data, it was assumed that people experienced moderate anxiety while waiting for a potential positive result for BCC. Moreover, it was assumed that people already had a utility of < 1, and that they moved from a state of no anxiety/depression to moderate anxiety/depression. In the health state valuation equation provided by Dolan62 for EQ-5D (shown in Diagnostic economic model on suspected melanoma lesions following an equivocal finding on dermoscopy: methods), the disutility (coefficient) for moderate depression/anxiety was –0.071. According to clinical expert advice, the results of diagnostic biopsy for suspected BCC are available 6 weeks after the procedure. Therefore, the total reduction in QALYs associated with the anxiety while waiting for the results of diagnostic biopsy for suspected BCC was estimated to be –0.008. This disutility was applied in every person waiting for results, regardless of the person’s number of lesions awaiting diagnosis.

A number of people may experience permanent disutility because of scars on their head or neck caused by diagnostic biopsy of surgical treatment of skin lesions. In the economic model it was assumed that 5% of people undergoing a diagnostic biopsy for a skin lesion on their head or neck and 15% of people undergoing surgery for BCC on their head or neck would experience permanent disutility because of their scar over their lifetime. Clinical expert advice was that 100% of diagnostic biopsies for suspected BCC are undertaken with simple repairs/scars; surgical excisions make up 75% of simple and 25% of complex repairs/scars, whereas in Mohs surgery simple and complex repairs/scars make up 50% each. Based on these estimates and the disutility data reported in Seidler et al. ,57 the permanent disutility from scarring following diagnostic biopsy, surgical excision and Mohs surgery was estimated to be –0.016, –0.019 and –0.021, respectively. These disutilities were applied only to people who experienced a permanent reduction in their HRQoL because of scarring on their head or neck.

It should be noted that, as the general utility of people was not expected to differ between the two arms of the economic model (apart from the disutilities described above associated with certain procedures and resulting scars), the total number of QALYs in each arm in the model, reflecting the overall utility of each model arm from start of the model and over lifetime, was not estimated. The mean number of QALYs reported for each arm of this model is therefore negative, and reflects only the total disutility experienced by each arm of the model because of the biopsy, surgery and/or scarring resulting in permanent disutility over the time horizon of the analysis.

Table 31 provides all utility data applied in the diagnostic economic model on lesions suspicious of BCC.

Costs

Costs considered in this economic model included the cost of diagnostic assessment with VivaScope following a positive result on dermoscopy, the cost of diagnostic biopsy and the cost of treatment (including the cost of unnecessary treatment for skin lesions with a FP result in VivaScope examination).

As reported in Table 24, the cost of VivaScope per suspected BCC lesion examined was estimated to be £71 if VivaScope is exclusively used for the diagnostic assessment of suspected BCC lesions found positive on dermoscopy; £62 if VivaScope is used only for the diagnostic assessment of suspected melanomas giving an equivocal finding on dermoscopy and suspected BCC lesions with a positive dermoscopic finding; and £58 if the device is used not only for the diagnostic assessment of suspected melanomas and BCCs, but also for the mapping of LMs prior to surgical treatment.

The costs of all other procedures and treatments included in the model were taken from either the NHS Reference Costs 2013 to 201412 for 2014 or the NHS national drug tariff for 2015. 64 Clinical experts advised on the appropriate NHS service and procedure codes corresponding to procedures and treatments considered in the model.

The unit cost of diagnostic biopsy was estimated to be £134, corresponding to the national unit cost of outpatient minor skin procedures conducted in a dermatology service for people aged ≥ 13 years (service code 330, currency code JC43A). 12

Treatment comprised a mixture of surgical and non-surgical therapies. Clinical experts indicated that the proportion of BCC lesions treated surgically ranges between 66% and 90% of BCC lesions. The economic model assumed that 75% of BCC lesions are treated surgically. Of those, 85% were assumed to undergo surgical excision and 15% to be treated with Mohs surgery (the proportion of BCC lesions undergoing Mohs surgery among those receiving surgical treatment appears to range between 10% and 20% across services in the UK, as indicated by clinical experts, although a wider variation may potentially exist).

Among lesions managed with non-surgical treatment, the percentage of lesions receiving each treatment was derived from a multicentre audit (seven centres in the Mersey region of north-west England), comprising a retrospective case note review of 50 randomly selected patients per trust who had BCCs managed non-surgically within a 12-month time period (1 January 2012–1 January 2013). 94 In total, 246 patients were selected as being suitable for the audit. The most commonly used agent for treatment was imiquimod (used by ≥ 50% of patients with BCC), followed by photodynamic therapy in 21%, radiotherapy in 19% and fluorouracil in 8%. Based on these data, and after consulting with clinical experts, it was assumed that non-surgical treatment of BCCs in the economic model comprised 60% topical treatment with imiquimod or fluorouracil (30% each), 21% photodynamic therapy and 19% radiotherapy. It should be emphasised that this is not necessarily a typical picture of non-surgical treatments across the country, as the EAG was advised that some of these treatments are not routinely used in some dermatology services, whereas others (such as cryotherapy, and curettage and cautery), which were not included in the economic model, may be more frequently offered in some services for the treatment of low-risk BCC lesions, especially superficial ones. However, regarding non-surgical therapies, as these made up only 25% of the treatment of BCC lesions, the impact of variations in relevant practice across settings on the total cost of BCC treatment was rather insubstantial.

By combining the above resource-use estimates with appropriate unit costs12,64 as recommended by clinical experts, the mean total cost of treatment per BCC lesion was estimated at £475.

All other health-care and PSS costs incurred by people in the model were estimated to be equal between the two arms of the model and were, thus, omitted from the analysis.

Table 32 provides the data and assumptions used at the estimation of the mean weighted treatment cost of BCC.

All input parameters utilised in the diagnostic economic model on lesions suspicious of BCC following a positive dermoscopic finding are shown later (see Table 35).

Study population

The study population for this model comprised patients with LM, aged 70 years, undergoing margin delineation prior to receiving surgical treatment. The aim of examination of LM with VivaScope prior to surgical removal was accurate definition of tumour margins. Surgical removal of LM needs to balance between sufficiently wide margins to prevent recurrence, and minimal margins to preserve functional and aesthetic areas of the face and neck. Therefore, accurate definition of the surgical margins of LM potentially leads to a low rate of multiple excisions, sparing tissue in functional and aesthetic areas. 95

Epidemiological data specific to LM are rather sparse in the literature, possibly because this is a precancerous condition and it may not always be recorded in cancer registries. Cases of LM are not routinely included in UK cancer statistics. LM is more common in older people. A review of LM and LMM, published in 1995, indicated that patients with LM are generally > 40 years of age, with a mean age of 65 years. 96 LM most commonly affects the sun-exposed skin of the head and neck, with a predilection for the cheek. 96 A recent US study identified all adult residents with a first lifetime diagnosis of LM between 1970 and 2007 in Olmsted County, MN, USA. 70 The study analysed medical records in order to determine demographic, clinical and surgical data, as well as incidence and survival rates associated with LM. 70 According to this study, the mean age of patients at LM diagnosis was 70 years (range 33–97 years), with 64.1% being male. The proportion of LMs on the head or neck was approximately 62%. However, clinical expert advice to the EAG indicated that this percentage might be much higher, and even reach 90%. Based on this information, the study population in the economic model had a mean age of 70 years, with 64% being male and 70% of them having a LM on the head or neck. Each person had only one diagnosed LM that required surgical treatment at the time of the analysis, according to clinical expert opinion.

The annual number of LMs examined for margin delineation at a dermatology MDT service in the UK was estimated to approximate 75, as reported in Annual number of cases eligible for examination with VivaScope in a dermatology multidisciplinary team clinic in the UK.

Intervention and comparator

The intervention assessed in this model was VivaScope 3000 for the margin delineation of LM prior to surgical treatment. The comparator was routine practice, which comprised presurgical assessment of LM margins with a dermoscope and/or clinical judgement.

Model structure

A decision tree followed by a Markov model was constructed to assess the cost-effectiveness of VivaScope in the margin delineation of LMs prior to surgical treatment. According to the model structure, which was determined by clinical expert advice and availability of relevant data, patients of 70 years of age with a LM planned for surgical treatment either had their tumour examined with VivaScope 3000 for margin delineation prior to surgery or underwent routine management, comprising presurgical assessment of LM margins with a dermoscope and/or clinical judgement.

Following margin assessment, LMs in both arms of the model were removed either by surgical excision or by Mohs surgery. A proportion of surgical excisions were incomplete, as determined by histopathology, meaning that some premalignant cells were still present after treatment, despite margin delineation. Incompletely excised tumours required a second surgical excision 4–6 weeks later, after which excision was assumed to be complete and confirmed by histopathology. The proportion of LMs that were incompletely excised was determined by the type of presurgical assessment of the margins (i.e. by VivaScope 3000 or dermoscopy/clinical judgement). Mohs surgery is performed in surgical stages until the surgical margins are clear. The type of presurgical assessment of the margins (i.e. by VivaScope 3000 or dermoscopy/clinical judgement) affected the number of stages of Mohs surgery.

All patients experienced distress because of surgery. Moreover, a proportion of patients with a LM surgically removed from their head or neck experienced a permanent reduction in their HRQoL because of the resulting scarring.

After complete surgical excision or Mohs surgery, all patients in both arms of the decision tree entered the Markov model, which was run in yearly cycles; half-cycle correction was applied. All patients entering the Markov model were at risk of recurrence of their tumour for the first 10 years (i.e. 10 years after the primary surgical removal of their LM). The risk of recurrence depended on the type of initial presurgical margin delineation (i.e. with either VivaScope 3000 or dermoscopy/clinical judgement) and/or the type of initial surgical treatment they had received (i.e. surgical excision or Mohs therapy). Patients experiencing a recurrence either underwent surgical excision or Mohs surgery, as according to clinical expert advice the vast majority of LMs are treated surgically; alternative therapies (such as radiotherapy and topical therapy with imiquimod) are used only if the patient is unfit for surgery or if there is a medical reason preventing surgery, for example if the patient is very frail and elderly. All patients experienced distress because of surgical treatment. A proportion of the patients with LMs on the face could experience permanent disutility because of scarring, if they were not already experiencing permanent disutility as a result of scarring after the initial surgical treatment.

The Markov model consisted of the states of ‘no recurrence, no permanent disutility because of scarring’, ‘no recurrence, permanent disutility because of scarring’, ‘recurrence, no permanent disutility because of scarring’, ‘recurrence, permanent disutility because of scarring’ and ‘death’, which was an absorbing state. Patients moving from the decision tree could enter any of the Markov model states (except death), depending on whether or not they had already experienced permanent disutility because of scarring. Patients in the ‘no recurrence, no permanent disutility because of scarring’ state could remain in this state, experience a recurrence and move to ‘recurrence, no permanent disutility because of scarring’ state, experience a recurrence and a scar that created permanent disutility thus moving to ‘recurrence, permanent disutility because of scarring’ state (this was possible for only patients with LM on the head or neck) or die. Patients in the ‘no recurrence, permanent disutility because of scarring’ state (who were patients with a LM on their head or neck) could remain on this state, experience a recurrence and move to ‘recurrence, permanent disutility because of scarring’ state or die. The two recurrence states with or without permanent disutility because of scarring were only temporary states; patients in these states could only transition to the two non-recurrence states with or without permanent disutility because of scarring, respectively, from which they could transition to a new recurrence or death in the next cycle. After the first 10 years, patients could not experience a recurrence of their tumour and therefore they could either remain in their ‘no recurrence’ state (with or without scarring) or die.

Lentigo maligna in the economic model was assumed not to progress to LMMs, as the relevant risk was low, given that all LMs in the model were treated.

The time horizon of the economic model was over a lifetime (up to 100 years of age). A schematic diagram of the VivaScope margin delineation model is shown in Figure 8.

FIGURE 8.

Structure of the margin delineation model. (a) Decision tree component; and (b) the Markov model component. a, As confirmed by histopathology. Note that dotted lines in the Markov model illustration show points of entrance from the decision tree.

Clinical input parameters

Utility values

Patients in this model experienced a reduction in their HRQoL for one of the following reasons:

• surgical treatment (either surgical excision or Mohs surgery)

• permanent scarring following surgical treatment of a LM on the head or neck.

As reported in Systematic literature review of existing economic evidence, Results, Seidler et al. 57 estimated a disutility of –0.004 associated with an excision procedure because of facial non-melanoma skin cancer using traditional surgical excision or Mohs surgery. They also reported a disutility of –0.016 for simple repairs/scars (granulation and primary closure) and a disutility of –0.026 for complex repairs/scars (local flap and graft).

The presence and surgical management of LM was considered to have a similar impact on patients’ HRQoL as the presence and surgical management of BCC. Owing to a lack of more relevant and better-quality data, the value of –0.004 was used to reflect the decrement in HRQoL (utility) experienced because of surgical treatment (either surgical excision or Mohs surgery). This disutility as a result of surgical treatment was applied as a one-off every time a person underwent surgical treatment (i.e. at first surgery, repeat surgery because of incomplete excision or future recurrence).

Disutility because of a patient waiting for a second surgery following incomplete excision was not considered; however, it is acknowledged that waiting time for a second surgery may create additional distress to the patient.

The number of stages in Mohs surgery is expected to affect the patients’ HRQoL, in terms of time and distress. However, the differential utility resulting from differences to the number of stages in Mohs surgery was not factored into the model as it was not possible to estimate a disutility per stage.

A number of people may experience permanent disutility because of scars on their head or neck as a result of the surgical removal of LMs. In the economic model it was assumed that 15% of patients undergoing surgical treatment for their LM on their head or neck (either for the first time or at a future recurrence of the tumour) would experience permanent disutility because of their scar. Clinical expert advice was that surgical removal of LMs, either by surgical excision or by Mohs surgery, is made up of 50% simple and 50% complex repairs/scars. Based on these estimates and the disutility data reported in Seidler et al. ,57 the disutility associated with scarring from surgical treatment of LMs was estimated to be –0.021. This disutility was applied only to people with permanent reduction in their HRQoL because of scarring on the head or neck. An individual who underwent surgical treatment for a LM on the head or neck and did not experience disutility because of scarring was at a 15% risk of experiencing permanent disutility caused by scarring at each potential future recurrence of LM.

As the general utility of people in the model was not expected to differ between the two arms of the economic model (apart from the disutilities described above associated with surgical treatment and resulting scars), the total number of QALYs in each arm in the model, reflecting the overall utility of each model arm from start of the model and over a lifetime, was not estimated. The mean number of QALYs reported for each arm of this model is, therefore, negative and reflects only the total disutility experienced by each arm of the model caused by surgery and/or scarring resulting in permanent disutility over the time horizon of the analysis.

Table 33 provides all utility data applied in the economic model on margin delineation of LMs.

Costs

Costs included the cost of presurgical mapping of LMs with either VivaScope 3000 or dermoscopy/clinical judgement, the cost of treatment with either surgical excision or Mohs surgery and the cost of potential future treatment because of recurrence.

As reported in Table 24, the cost of margin delineation with VivaScope 3000 per LM mapped was estimated to be £250 if VivaScope is exclusively used for presurgical margin delineation of LMs and £105 if the device is used for the diagnostic assessment of suspected melanomas and BCCs as well as for the mapping of LMs prior to surgical treatment.

Routine presurgical margin delineation of LMs with dermoscopy/clinical judgement was estimated to comprise 5 minutes of a consultant dermatologist’s time. Using the unit cost of a consultant dermatologist of £140 per hour of contract,65 the mean cost of routine presurgical margin delineation was estimated to be £12. The acquisition cost of dermoscopy was not included in the estimation of the cost of routine presurgical margin delineation LMs, as dermoscopes appear to be already in place in dermatology departments and can be used for the assessment of skin lesions.

The proportion of LMs that were treated with surgical excision in the first surgery following margin delineation and in future recurrences was estimated based on clinical expert opinion. It was assumed that 85% of the first surgical treatment of LMs made up surgical excision and 15% Mohs surgery. After tumour recurrence, it was assumed that 80% of LMs were treated with surgical excision and 20% with Mohs surgery.

The unit costs of surgical excision and Mohs surgery were taken from the NHS reference costs for 2014. 12 Clinical experts advised on the appropriate NHS service and procedure codes corresponding to these two types of surgical treatment.

Mohs surgery is undertaken in stages. The number of stages required for Mohs surgery is directly related to an opportunity cost in terms of staff time and consumables; however, it was not possible to identify a unit cost per stage of Mohs surgery. For this reason, it was assumed that the unit cost reflecting Mohs surgery corresponded to the mean number of required stages, across all skin operations requiring Mohs surgery. As it was not possible to identify relevant UK data on the mean number of stages required in Mohs surgery, this figure was derived from a US multicentre prospective cohort study that aimed to evaluate the rate of complications and postoperative pain associated with the treatment of skin cancer using Mohs surgery. 102 The study included 1550 patients with 1792 tumours, the majority of which were BCC (61%) or SCC (31%). The authors reported that the mean number of stages was 1.6, ranging from 1 to 8. Therefore, for the purposes of costing, it was assumed that the national unit cost reflecting the cost of Mohs surgery corresponded to 1.6 Mohs stages, and that 70% of this unit cost was fixed (and independent of the number of stages involved in Mohs surgery), whereas the remaining 30% of the unit cost was variable and in a linear relationship with the number of stages required for the completion of Mohs surgery. This assumption was utilised only in the first surgery following margin delineation of LMs. For Mohs surgery undertaken in future recurrences, the mean cost of Mohs surgery, without adjusting for the number of stages, was used.

All other health-care and PSS costs incurred by people in the model were estimated to be equal between the two arms of the model and were thus omitted from the analysis.

Table 34 provides the percentages of each type of surgical treatment for LM at first surgery following margin delineation and after recurrence, the costs of each type of surgical treatment, as well as the data sources and assumptions used for their estimation.

All input parameters utilised in the diagnostic economic model on lesions suspicious of BCC following a positive dermoscopic finding are shown later (see Table 35).

Overview of methods of analysis

A deterministic analysis that utilised point estimates of each model input parameter was first undertaken. This was followed by a probabilistic analysis, which was conducted to take account of the uncertainty characterising the input parameter estimates. For this analysis, all relevant input parameters were entered as probability distributions to reflect their imprecision. Probability distributions were determined by the available data or, where data were lacking, by plausible assumptions. Monte Carlo simulation was then employed to reflect this uncertainty in the models results: 10,000 iterations were performed, each drawing random values out of the distributions fitted onto the model input parameters. Results of the probabilistic analysis were averaged across the 10,000 iterations to provide a mean estimate of costs and QALYs for each intervention. In addition, uncertainty in the model input parameters and structural assumptions were explored through deterministic one-way and two-way sensitivity analyses.

The results have been presented in the form of incremental cost-effectiveness ratios (ICERs), except in cases of dominance, which occurs when an intervention results in lower costs and a higher number of QALYs than its comparator. The results of both types of analyses (deterministic and probabilistic) have been depicted in the form of cost-effectiveness planes. The results of the probabilistic analysis have been summarised in the form of cost-effectiveness acceptability curves (CEACs), which show the probability of VivaScope being cost-effective at different cost-effectiveness thresholds, in each of the analyses considered. All input parameters were tested in one-way sensitivity analysis; tornado diagrams were produced for different analyses to show the impact of the most influential parameters on the results. The results of tornado diagrams have been reported using incremental net monetary benefits (INMBs), estimated at a willingness-to-pay (WTP) threshold of £20,000/QALY, rather than ICERs, because use of the whole range of some parameters tested in the tornado diagrams resulted in negative ICERs, because of dominance, which are not meaningful. Additional one-way sensitivity analyses were undertaken to estimate the impact of alternative scenarios and model assumptions on the results. Finally, two-way sensitivity analyses were carried out to test the impact of concurrently varying sensitivity and specificity of VivaScope in the diagnostic assessment of eligible skin lesions suspicious of melanoma or BCC on the cost-effectiveness results.

Summary of all model input parameters, probability distributions and range of values tested in sensitivity analysis

In order to run the probabilistic analysis, all relevant input parameters were entered as probability distributions to reflect their imprecision. Probability distributions were determined by the available data or, where data were lacking, by plausible assumptions.

The annual number of the three types of lesions examined with VivaScope (i.e. suspected melanomas with an equivocal finding on dermoscopy, suspected BCCs with a positive or equivocal dermoscopic finding and LMs undergoing presurgical margin delineation) was given a uniform distribution, with a range of ± 30% of the originally estimated volume.

The diagnostic accuracy characteristics of VivaScope and monitoring (which was part of routine management of equivocal lesions suspicious of melanoma) (i.e. that is sensitivity and specificity) were given a beta distribution. It is acknowledged that sensitivity and specificity are usually correlated, and, as such, a joint distribution should ideally be used. However, as no meta-analysis of diagnostic accuracy data was performed and no summary receiver operating characteristic curves that could indicate the relationship between sensitivity and specificity were possible to produce, as described in Chapter 3, it was considered reasonable to use a beta distribution for sensitivity and specificity, assuming that these are independent from each other, although this assumption is acknowledged as a limitation of the analysis.

All proportions and dichotomous probabilities (e.g. the proportion of men in the study population, the proportion of lesions on the head or neck, the probability of death associated with melanoma, the prevalence of cancer in lesions suspicious of skin cancer, the probability of future recurrence of LM) were given a beta distribution. Utilities were also given a beta distribution, using the method of moments; disutilities were given a distribution of 1 minus beta. Polychotomous transitions and variables were given a Dirichlet distribution.

Staff unit costs (radiographer, consultant dermatologist) and the required staff time to operate the VivaScope system were given a normal distribution. All other costs were assigned a gamma distribution.

Table 35 provides an overview of all input parameters, reporting deterministic values and details on the types and range of probability distributions assigned to each parameter with relevant data sources and justifications.

Table 36 provides the mean and the range of values of the most influential model input parameters depicted in tornado diagrams, together with a justification of the extreme values used for each parameter.

Additional scenarios tested in one-way sensitivity analysis

Further to tornado diagrams, which depicted the impact of the most influential input parameters on the results of the economic analysis, additional sensitivity analyses were carried out to explore the robustness of the results under alternative scenarios and model assumptions. The following alternative scenarios were explored.

Relating to the cost of VivaScope examination:

• The estimated staff time cost for the diagnosis of skin lesions suspicious of cancer was replaced by the NHS reference cost of £47 for a direct-access ultrasound scan of < 20 minutes, as a proxy;12 the estimated staff time cost for mapping of skin lesions prior to surgical treatment was replaced by the NHS reference cost of £109 for a consultant-led, outpatient, dermatology first visit. 12

• The cost associated with training was doubled, to account for the extra training required over the first few months in order for dermatologists to gain experience in the clinical interpretation of the results obtained from the examination of lesions with VivaScope. In addition, the useful time of training was reduced to 5 years.

Relating to the diagnostic model on suspected melanomas:

• People waiting for the results of biopsy were assumed to experience moderate rather than severe anxiety; therefore, a much lower disutility of anxiety of –0.071 was used in this scenario, as estimated from the health state valuation equation provided by Dolan62 for EQ-5D, rather the value of –0.505 that was used in the base-case analysis.

Relating to the diagnostic model on suspected BCCs:

• Clinical experts advised that, in reality, not all suspected BCCs receive diagnostic biopsy following dermoscopy, but some move on directly to treatment. Therefore, a scenario was tested where only 70% of suspected BCCs received a diagnostic biopsy under routine care; that is, only the 70% of the diagnostic biopsy cost was applied and only 70% of people were assumed to experience disutility associated with biopsy and permanent scarring following biopsy on the head or neck (unless surgical treatment was received). For simplicity, it was assumed that the percentage of 70% of suspected BCCs that received diagnostic biopsy did not distinguish between true BCCs and no BCCs; in other words, both suspected BCC lesions that proved to be BCCs and suspected BCCs that were not actually BCCs were subject to a 0.7 probability of biopsy under this scenario.

In addition to the above scenarios, the ICERs obtained in each model were plotted against different values of the annual number of each type of lesion examined with VivaScope (i.e. equivocal lesions suspicious of melanoma, suspected BCCs that give a positive or equivocal finding on dermoscopy and LMs prior to surgical treatment) to identify the minimum number of each type of lesion that is required to be examined with VivaScope per year, so that examination with VivaScope is a cost-effective strategy.

Finally, for the diagnostic model on suspected melanomas that utilised diagnostic accuracy data from Pellacani et al. ,42 the impact of the percentage of the suspected melanomas with an equivocal finding on dermoscopy that were excised (i.e. because they were highly suspicious) on the results was assessed by plotting the ICER obtained in the respective analysis against the whole range of the probability of suspected melanomas being excised (i.e. 0–100%). This was decided because the percentage of the suspected melanomas that were excised on the results of the analysis had a twofold impact:

1. An increase in the percentage of suspected melanomas that were excised led to a lower diagnostic accuracy of VivaScope examination, as Pellacani et al. 42 reported a lower specificity for VivaScope in lesions that were chosen for excision as highly suspicious following dermoscopy than the specificity of VivaScope in less suspicious lesions that were selected for monitoring based on the results of dermoscopy. Consequently, an increase in the percentage of suspected melanomas being excised reduced the benefit of VivaScope in the model.

2. At the same time, an increase in the percentage of suspected melanomas that were excised led to an increase in the cost of routine management (as the cost of excision is higher than the cost of monitoring) and an increase in the disutility because of excision, permanent disutility because of scarring (relevant to lesions on the head or neck) and disutility because of anxiety while waiting for the results. Consequently, an increase in the percentage of suspected melanomas that were being excised increased the cost of routine management and reduced its benefit.

Base-case deterministic and probabilistic results

The base-case deterministic and probabilistic results for each of the economic models and analyses considered for this report are provided in Tables 37–46. For each type of lesion, different cost and cost-effectiveness results are presented, depending on the types of lesions expected to be examined with VivaScope; the latter determined the cost of VivaScope, as the total cost associated with acquisition and use of the device was spread across the annual number of lesions examined with VivaScope in order to determine a cost per lesion.

Results of the diagnostic model of suspected melanomas are presented in Tables 37 and 38 (results derived when diagnostic data from Alarcon et al. 30 were utilised) and in Tables 39 and 40 (results derived when diagnostic data from Pellacani et al. 42 were utilised). It can be seen that under use of the more optimistic diagnostic data from Alarcon et al. ,30 VivaScope appears to be cost-effective in the diagnostic assessment of suspected melanomas with an equivocal finding on dermoscopy, even when VivaScope is exclusively used for this purpose (with an ICER of £8877 per QALY in deterministic analysis and £9362 per QALY in probabilistic analysis). On the other hand, use of the diagnostic data from Pellacani et al. 42 resulted in an ICER of £19,095 per QALY in deterministic analysis and £25,453 per QALY in probabilistic analysis when VivaScope was considered for only the diagnostic assessment of equivocal lesions suspicious of melanoma. Nevertheless, if VivaScope is expected to be used in the diagnostic assessment of both suspected melanomas and suspected BCCs, or also in the mapping of LMs prior to surgical treatment, then VivaScope becomes dominant in the diagnostic assessment of melanomas, as the cost associated with its use is spread across a larger number of sessions, leading to the total cost associated with VivaScope examination in the economic model being lower than the total cost associated with routine management of equivocal lesions suspicious of melanoma.

VivaScope was shown to be the dominant strategy when used for the assessment of suspected BCCs, regardless of its estimated use exclusively for this purpose or for the assessment of suspected melanomas and LMs as well (see Tables 41 and 42). Consideration of the use of VivaScope for other indications, further to its use on the diagnostic assessment of suspected BCCs, had little impact on the results as the annual number of suspected BCCs is much higher than the annual number of other lesions expected to be examined with VivaScope and, therefore, this number of suspected BCCs drives the cost per lesion examined with BCC and, subsequently, the cost-effectiveness results.

Regarding margin delineation of LMs, mapping with VivaScope was shown to be cost-effective, even if it used exclusively for this purpose, as indicated by an ICER of £10,241 per QALY obtained in deterministic analysis (see Table 43) and £11,651 per QALY in probabilistic analysis (see Table 44). When use of VivaScope was expanded to other indications covered in this economic analysis, then VivaScope became the dominant option.

Overall, in the analyses that combined the different ‘part’ models designed for this report, VivaScope was shown to be the dominant strategy over routine management in the diagnostic assessment of suspected melanomas and BCCs (see Table 45) and in the diagnostic assessment of suspected melanomas and BCCs combined with margin delineation of LMs prior to surgical treatment (see Table 46). The tables show the deterministic results, but probabilistic results were very similar.

The cost-effectiveness planes of all the probabilistic analyses undertaken for this assessment are provided in Appendix 6.

The CEACs for each part model considered in the analysis are provided in Figures 9–12. Figure 9 indicates that, using the diagnostic accuracy data from Alarcon et al. ,30 the probability of VivaScope being cost-effective in the diagnostic assessment of suspected melanomas is zero at a zero WTP per QALY gained, but reaches 0.99 at the lower NICE cost-effectiveness threshold of £20,000 per QALY when VivaScope is used only for this purpose (i.e. diagnostic assessment of suspected melanomas). When VivaScope is used for the diagnostic assessment of suspected melanomas and BCCs or a combination of diagnosis of suspected melanomas and BCCs and presurgical margin delineation of LMs, then the probability of being it cost-effective in the diagnosis of suspected melanomas is 1 and is independent of the level of WTP considered.

FIGURE 9.

Cost-effectiveness acceptability curve: diagnostic model on suspected melanomas – diagnostic data based on Alarcon et al. 30

FIGURE 10.

Cost-effectiveness acceptability curve: diagnostic model on suspected melanomas – diagnostic data based on Pellacani et al. 42

FIGURE 11.

Cost-effectiveness acceptability curve: diagnostic model on suspected BCCs.

FIGURE 12.

Cost-effectiveness acceptability curve: presurgical margin delineation model on LMs.

Figure 10 shows the CEAC derived when using the diagnostic accuracy data for suspected melanomas from Pellacani et al. 42 In this case, the probability of VivaScope being cost-effective when used exclusively in the diagnostic assessment of suspected melanomas is 0.29 and 0.69 at the lower and upper NICE cost-effectiveness threshold, respectively. When the use of VivaScope is expanded to the diagnostic assessment of suspected BCCs or all indications examined in this analysis, its probability of cost-effectiveness in the diagnostic assessment of suspected melanomas reaches 0.99 at the NICE lower cost-effectiveness threshold of £20,000 per QALY.

Regarding the probability of cost-effectiveness of VivaScope in the diagnostic assessment of suspected BCCs, Figure 11 shows that this is 1, regardless of whether VivaScope is used exclusively for this purpose or its use is expanded to other indications examined in this economic evaluation, and is independent of the cost-effectiveness threshold used.

Finally, Figure 12 provides the CEAC for the model assessing the cost-effectiveness of VivaScope in the presurgical margin delineation of LMs. It shows that when VivaScope is used exclusively for this purpose, its probability of being cost-effective is 0.62 and 0.74 at the lower and upper NICE cost-effectiveness threshold, respectively. However, when VivaScope is used for all indications considered in economic modelling, its cost-effectiveness in the presurgical margin delineation of LMs improves, and its probability of being cost-effective rises up to 0.92 and 0.94 at the lower and upper NICE cost-effectiveness threshold, respectively.

Results of one-way and two-way sensitivity analyses

One-way sensitivity analyses were performed on all input parameters that were given a probability distribution in the economic model. The tornado diagrams that present the impact of the most influential input parameters on the results are shown in Appendix 6. It is evident that among the most influential parameters across all models are those relating to permanent disutility because of scarring (such as the percentage of people experiencing permanent disutility as well as the value of disutility itself) and the disutility because of anxiety while waiting for the results of biopsy. Overall, the most influential parameters included the following points.

In the diagnostic assessment of suspected melanomas:

• the percentage of people experiencing permanent disutility because of scarring

• the disutility because of anxiety while waiting for the biopsy results

• the percentage of equivocal lesions excised under routine management (this parameter was not considered in the tornado diagrams when Pellacani et al. 42 data were used, because of its twofold impact on the results, which would lead to a misleading picture in the tornado diagram)

• the permanent disutility because of scarring from first excision

• the annual number of suspected melanomas eligible for examination for VivaScope (if VivaScope was used exclusively for examination of suspected melanomas)

• the VivaScope sensitivity and specificity

• the prevalence of melanomas in equivocal lesions

• the cost of first excision

• the disutility because of the first excision.

It should be noted that when VivaScope was assumed to be used exclusively for the diagnosis of suspected melanomas and diagnostic data from Alarcon et al. 30 were utilised, the only parameter that potentially resulted in negative INMBs in the tornado diagram was the disutility because of anxiety. When VivaScope was assumed to be used exclusively for the diagnosis of suspected melanomas and diagnostic data from Pellacani et al. 42 were utilised, then several parameters resulted in negative INMBs. However, when use of VivaScope was assumed to expand to diagnosis of suspected BCCs as well, none of the influential parameters could result in a negative INMB. Tornado diagrams were not produced for the scenario of VivaScope being used for all indications suggested in this economic analysis, as results were expected to be similar to those produced when diagnosis of both suspected melanomas and suspected BCCs was informed by VivaScope.

In the diagnostic assessment of suspected BCCs:

• the percentage of people experiencing permanent disutility because of scarring from biopsy

• the disutility because of anxiety while waiting for the results

• the diagnostic biopsy cost

• the prevalence of BCC in examined lesions

• the permanent disutility because of scarring from biopsy

• the annual number of suspected BCCs that would be examined with VivaScope

• the disutility because of the biopsy

• the percentage of patients treated with surgical therapy

• the sensitivity of VivaScope 3000

• the number of lesions per person

• the percentage of people experiencing permanent disutility because of scarring from surgery.

However, none of the parameters had such an impact so as to turn the INMB to negative values, even when VivaScope was used exclusively in the diagnostic assessment of suspected BCCs. For this reason, tornado diagrams relating to expansion of use of VivaScope for the assessment of other types of lesions were not produced, as expansion of use of VivaScope would only reduce the impact of influential parameters on the results even further.

In the presurgical mapping of LMs:

• the probability of incomplete surgical excision following routine mapping

• the probability of annual recurrence after surgical excision

• the probability of incomplete surgical excision following mapping with VivaScope

• the permanent disutility because of scarring from surgical treatment

• the percentage of people with permanent disutility from scarring

• the probability annual recurrence following VivaScope mapping and surgical excision

• the VivaScope mapping (staff) time

• the cost of surgical excision

• the number of Mohs stages under routine mapping

• the disutility caused by surgery.

As with the results for suspected melanomas, a number of influential parameters could turn the INMB into a negative value if VivaScope was used only for the mapping of LMs prior to surgical treatment. However, when a wider use of VivaScope was assumed, the INMB remained positive under any values of the influential parameters examined.

The results of the additional sensitivity analyses are shown in Table 47. It can be seen that results for suspected melanoma are negatively affected after application of relevant scenarios, when diagnostic accuracy data from Pellacani et al. 42 are used and VivaScope is assumed to be exclusively used for the diagnostic assessment of suspected melanomas. However, when wider use of VivaScope is assumed, the results are practically unaffected by the scenarios tested.

Two-way sensitivity analyses were performed to test the impact of different combinations of sensitivity and specificity of VivaScope on its cost-effectiveness in the diagnostic assessment of equivocal lesions suspicious of melanoma. The results on the diagnosis of suspected melanomas are shown in Tables 48 and 49. The results indicate that VivaScope needs to have a relatively high diagnostic accuracy in order to be cost-effective, in particular when it is used exclusively for the diagnostic assessment of suspected melanomas. A two-way sensitivity analysis for the diagnosis of suspected BCCs showed that any combination of sensitivity and specificity from values as low as 0.40 resulted in VivaScope being a cost-effective strategy (the maximum ICER, when sensitivity and specificity were 0.40, was £7083/QALY).

Figures 13–15 show the ICERs obtained in each model plotted against different values of the annual number of each type of lesion examined with VivaScope and help identify the minimum number of each type of lesion required to be examined with VivaScope per year, so that examination with VivaScope is a cost-effective strategy. For suspected melanomas and LMs only, exclusive use of VivaScope for their examination is shown in the graphs, because consideration of wider use of VivaScope resulted in VivaScope being dominant in the diagnosis of suspected melanomas and mapping of LMs, even when a negligible number of lesions examined (close to zero) was assumed.

FIGURE 13.

Incremental cost-effectiveness ratio plotted against annual number of suspected melanomas examined with VivaScope: exclusive use of VivaScope for this purpose. (a) Using data from Alarcon et al. ;30 and (b) using data from Pellacani et al. 42

FIGURE 14.

Incremental cost-effectiveness ratio plotted against annual number of suspected BCCs examined with VivaScope: different uses of VivaScope considered.

FIGURE 15.

Incremental cost-effectiveness ratio plotted against annual number of LMs mapped with VivaScope prior to surgical treatment: exclusive use of VivaScope for this purpose.

Finally, Figure 16 shows the impact of a change in the percentage of equivocal lesions suspicious of melanoma that are excised under routine management. The shape of the line is determined by the fact that the percentage of equivocal lesions sent for excision affects both the cost and disutility of routine management, but also the diagnostic accuracy of VivaScope, which differs between highly suspicious and low/moderately suspicious lesions in Pellacani et al. 42 The ICER is below the lower NICE cost-effectiveness threshold of £20,000/QALY when the percentage of equivocal lesions excised is approximately ≤ 10% or ≥ 60%.

FIGURE 16.

Incremental cost-effectiveness ratio plotted against the percentage of equivocal lesions suspicious of melanoma that are excised under routine management (highly suspicious lesions): diagnostic data from Pellacani et al. 42

Summary of key results of clinical effectiveness

Generalisability of results

Although none of the included studies in the review of clinical effectiveness was conducted in the UK, two studies30,42 on diagnosis and one study on margin delineation35 were deemed to be the most representative of clinical practice in the UK setting. Our clinical experts validated this and these trials were taken forward for the health economic analysis.

Strengths

• This systematic review provides the most up-to-date evidence of the clinical effectiveness of VivaScope 1500 and 3000 for detecting and monitoring skin cancer, and with a low likelihood of missing any key or pivotal trial.

Limitations

• There is lack of UK data among the included studies and, therefore, the generalisability of the results is limited. This has implications for the NHS.

• Apart from diagnostic accuracy and lesion recurrence rate (only reported by one study), none of the outcomes specified in the protocol was reported in the included studies.

• None of the included studies reported diagnostic accuracy results of SCC with VivaScope. This confirms evidence in the literature which suggests that SCCs can be difficult to view using imaging techniques because their upper surface is often scaly, which can make it difficult to obtain detail at sufficient resolution. 11 SCC will, therefore, not be carried through into the economic evaluation.

• In some of the studies, there was paucity and/or quality of reported data on number of patients with positive and negative test results, making it impossible to construct a 2 × 2 contingency table to calculate sensitivity and specificity.

Strengths

• The economic analysis was based on the development of three ‘part’ models, each designed to simulate the care pathways of people with skin lesions eligible for examination with VivaScope who undergo assessment of their skin lesions in a dermatology MDT service. The care pathways were designed based on national guidelines and following advice from clinical experts, and were specific to each type of lesion considered in the economic analysis. The use of national guidance and consultation with clinical experts ensured that the care pathways considered in this model reflect, as close as possible, clinical practice in the NHS, although there appears to be wide variation in the management of suspected and/or confirmed skin cancer across services.

• The model input parameters were based on national guidelines and other published evidence, clinical expert opinion and national unit costs.

Limitations

• The diagnostic and mapping accuracy data that were utilised in the model were taken from studies included in the systematic literature review of clinical evidence conducted for this guideline. However, data were limited and it was not possible to synthesise the results in a meta-analysis because of the heterogeneous nature of the studies identified. Moreover, none of the studies was conducted in the UK, which may have implications for the generalisability of not only the clinical, but also the economic, findings as the prevalence of the skin cancer and the population phenotype distribution may affect the diagnostic accuracy of VivaScope.

• Sensitivity analysis showed that the most influential parameters across all models were those relating to permanent disutility because of scarring following surgical intervention of skin lesions on the head or neck (such as the percentage of people experiencing permanent disutility as well as the value of disutility itself) and the disutility because of anxiety while waiting for the results of biopsy. However, utility data relating to these events were very limited and of poor quality or non-existent, and a number of assumptions were needed in order to inform the model.

• Other complications of excision and biopsy, which were the main comparators of VivaScope in the diagnostic assessment of suspected cancerous lesions, such as bleeding, bruising, infection or allergic reaction to the topical antibiotic were not considered in the model. Clinical experts acknowledged that these are not common complications, but their omission may have potentially underestimated, to some extent, the cost-effectiveness of VivaScope.