Health utilities for non‐melanoma skin cancers and pre‐cancerous lesions: A systematic review

Abstract Background Non‐melanoma skin cancers (NMSCs) are common and consume many healthcare resources. A health utility is a single preference‐based value for assessing health‐related quality of life, which can be used in economic evaluations. There are scarce data on health utilities for NMSCs. Objectives Using a systematic review approach, we synthesized the current data on NMSC‐related health utilities. Methods A systematic review of studies of NMSC‐related health utilities was conducted in Medline, Embase, and Cochrane databases. Data were extracted based on the protocol and a quality assessment was performed for each study. Results The protocol resulted in 16 studies, involving 121 621 participants. Mean utility values across the studies ranged from 0.56 to 1 for undifferentiated NMSC, 0.84 to 1 for actinic keratosis, 0.45 to 1 for squamous cell carcinoma, and 0.67 to 1 for basal cell carcinoma. There was considerable variability in utilities by type of cancer, stage of diagnosis, time to treatment, treatment modality, and quality of life instrument or method. Utility values were predominantly based on the EuroQol 5‐dimension instrument and ranged from 0.45 to 0.96, while other measurement methods produced values ranging from 0.67 to 1. Lower utility values were observed for advanced cancers and for the time period during and immediately after treatment, after which values gradually returned to pre‐treatment levels. Conclusions Most utility values clustered around relatively high values of 0.8 to 1, suggesting small decrements in quality of life associated with most NMSCs and their precursors. Variability in utilities indicates that careful characterization is required for measures to be used in economic evaluations.


| INTRODUCTION
Non-melanoma skin cancers (NMSCs), comprizing mostly basal cell carcinomas (BCCs) and squamous cell carcinomas (SCCs), are the most common cancers in countries with fair-skin populations. 1 In Australia, clinical management of BCCs and SCCs accounted for 8.1% of Australian health system spending on cancer in 2008-2009 (excluding screening). 2 The incidence of BCCs and SCCs is increasing in many countries, and the caseload is further impacted by ageing populations. [3][4][5][6] NMSCs are often under-reported as many cancer registries do not routinely gather this information 7,8 or the data collected is not separated by NMSC subtypes. 9 While most NMSCs (except Merkel cell carcinoma) are associated with a lower mortality rate than melanoma, 2 mortality rates for SCC are higher when diagnosed at a later stage, [10][11][12] therefore, in principle, early detection of skin cancer is beneficial. A higher recurrence rate of BCC and SCC and higher metastatic rate for SCC is generally expected for larger tumour size at diagnosis or those with high-risk features. 10,13 While uncomplicated and small tumours may be treated by cryotherapy, cautery, curettage or excision, more advanced cancers may require specialized procedures such as advanced reconstructive surgery, radiotherapy, chemotherapy and immunotherapy. 2,14 The potential benefits of a proposed screening or treatment should be balanced against the healthcare harms and economic costs, including those arizing from biopsies and surgical procedures, to determine the best use of limited healthcare resources. This is particularly pertinent for most NMSCs for which the mortality rates are relatively low, as benefits are unlikely to be reflected in reduced mortality but rather in improved quality of life.
Ascertaining health-related quality of life (HRQoL) is important to evaluate patient perspectives of benefits and harms of detection and treatment, which may include scarring, pain, post-operative complications and worry. 15 The domains of HRQoL are usually measured by generic instruments such as the EuroQol 5-dimension instrument 16 (EQ-5D), or dermatologyspecific instruments, such as Skindex, 17 the Basal and Squamous Cell Carcinoma QoL (BaSQoL) questionnaire, 18 and the Dermatology Life Quality Index (DLQI). 19 As a component of assessing HRQoL, it is often useful to generate a single preference-based value, called a health utility, combining all aspects of a health state that is measured on a scale from 0 (corresponding to death) to 1 (best possible health). The methods for obtaining health utilities are usually either direct, such as the time trade-off and standard gamble methods, 20 in which study participants indicate their preferences from a range of health states or scenarios, 21 or indirect, using HRQoL instruments, such as the EQ-5D. 16 Additionally, some non-preference based domains of HRQoL can be mapped to health utility scores using a validated algorithm. 22,23 Economic evaluations of healthcare programs commonly use health utilities to obtain quality-adjusted life years (QALYs), a measure that adjusts survival time by the quality of life within a health state. 24,25 An advantage of this approach is the ability to objectively assess differences in costs and QALYs across healthcare interventions, allowing policy-makers to compare the cost-effectiveness of healthcare programs to inform resource allocations, across different diseases. 24 Currently, there are limited data on health utilities for NMSCs. 26 This systematic review was undertaken to provide synthesized information for future health economic evaluations and to highlight the gaps in the current evidence.

| Search strategy
A search strategy was developed and encompassed literature on NMSCs and HRQoL. NMSCs were searched using general keywords and keratinocyte cancer as well as specific keywords for SCC, BCC, actinic keratosis (AK) and solar keratosis (Table S1, Appendix S1). Studies that included rarer NMSCs (such What is already known about this topic? � Non-melanoma skin cancers (NMSCs) are common and consume many healthcare resources. Health utilities are often used in economic evaluations of healthcare programs, however, there are scarce data on health utilities for NMSCs.

What does this study add?
� The systematic review suggests that most utility values clustered around relatively high values of 0.8 to 1, suggesting small decrements in quality of life associated with most NMSCs and their precursors. There was substantial variability in the utilities depending on cancer subtypes, treatments and methods, therefore, developing a comprehensive catalogue of NMSC health utilities is crucial for future health economic evaluations to inform policy and resource allocation decisions related to skin cancer early detection and management.
as Merkel cell carcinoma) were included if identified in the search but we did not include specific keywords for them. Keywords and MeSH terms were incorporated into the search (Table S1, Appendix S1). MEDLINE, Embase, and Cochrane Database of Systematic Reviews were searched from inception to April 2020. Results were limited to English language. The reference lists of relevant articles were hand-searched. The grey literature was not searched.

| Study selection
The inclusion criteria were full-text research articles: (1) referring to patients with cutaneous NMSCs, specifically BCC, SCC (including Bowen's disease or intraepithelial carcinoma) and their precursors (e.g., AK), and other rarer NMSCs such as Merkel cell carcinoma, and (2) evaluating health utilities of patients using direct or indirect methods. Although AK is not considered malignant, it was included in the search due to being a clinically detectable precursor lesion of SCC. 27 Authors of relevant conference abstracts were contacted to obtain full text or data. Titles and abstracts were initially screened (by CS) and full text articles of the short-listed studies were examined (by CS) with reference to the inclusion criteria.
Data extraction was independently performed (by CS and CW). Data extracted included: study population and setting, number of participants, mean age and gender, the instrument or method used to measure quality of life, utilities and any additional measures reported. Where treatment was specified, the subgroup numbers, treatment administered and the temporal relationship between treatment delivery and utility measurement were recorded.
Quality analysis of the studies was based on published guidelines for the conduct of systematic reviews of health state utility values. 28,29 The ROBINS-I risk of bias tool 30 was used to assess for bias due to missing data, as this is a common issue in quality of life studies. Where primary outcome data were not available, the authors were contacted for clarification.
The systematic review protocol was registered on the Prospero database (CRD42020179776). The PRISMA statement 31 was followed for conducting and reporting the study.

| Study characteristics
The initial search identified 589 studies, of which 116 were included from abstracts; 93 underwent full-text assessment ( Figure S1). Five additional studies were identified through the reference lists of relevant studies. The main reason for exclusion (120 studies) was the use of instruments from which utilities could not be derived, with the Skin Cancer Index, DLQI, and Skindex being the most common of such instruments.
BCC, SCC and AK were the most common conditions with some studies reporting pooled utility values from different types of NMSC without distinguishing the case-mix. Only one study 36 specifically referred to Bowen's disease (SCC in situ), and no studies considered Merkel cell carcinoma.
Five studies 33,39,40,44,46 were conducted in, or used treatments suitable for use in the community, and five studies 32,34,41,45,47 were in hospital settings. Four studies 35,36,42,43 drew participants from both the community and hospitals. Two studies were national population-based studies. 37,38 One study was a phase III clinical trial. 33 Out of the 15 studies that reported mean utilities, six 32,33,35,37,38,41 also reported median values (which were greater than the mean in all subgroups of four of the studies 32,33,35,41 ), one 36 also used assigned or calculated utilities, and seven 33,34,40,41,44,45,47 also reported standard deviations. One study 42 only reported median utility values.

| Quality of studies and risk of bias
Six studies had a population sample size of 100 or less. For risk of bias due to missing data, 30 11 studies were deemed low risk, two were moderate risk, two were serious risk, and one had insufficient information for assessment (Table S2). Studies that were labelled as moderate or serious risk for missing data had low response or participation rates.

| Categorization by skin cancer group and treatment
Studies that reported pooled utility values without distinguishing the types of skin cancer were categorized as NMSC, otherwise they were grouped by type of skin cancer.

| Undifferentiated non-melanoma skin cancer
Nine studies obtained health utilities for NMSC (Table 2). 32,34,35,37,38,40,41,43,46 The mean utility ranges were 0.72 to 1 and 0.56 to 1 for untreated NMSC (or where treatment was unspecified) and treated NMSC, respectively. Two studies 32,34 included patients with melanoma in their analysis, however the proportion of melanomas was relatively low (range 9%-16%). The NMSC utilities for these two studies ranged from 0.56 to 0.89. Two studies 32,34 analysed cancers of specific anatomical locations using EQ-5D with utility values of 0.56-0.65 (treated) for the lower limb and 0.72 (untreated) for the head and neck.
One study 43 focussed on cancers of the face and ears and obtained utilities using the time trade-off method of 0.97-1 (treated). Utilities for patients with lesions of suspected but unconfirmed malignant potential ranged from 0.97 to 0.98. 41

| Actinic keratosis
Six studies provided health states for AK ( Table 3). 33,35,39,41,42,46 Mean utility ranges were 0.84 to 0.99 and 0.89 to 0.96 for untreated AK (or where treatment unspecified) and treated AK, respectively. The utility for patients with pruritis was 0.92 (SD = 0.15), 41 which was included due to pruritis being a common side effect of topical treatments for AK.

| Squamous cell carcinoma
Health states pertaining to SCC were examined in four studies (Table 4). 35

| Basal cell carcinoma
Six studies examined health states for BCC (Table 5). 35,36,41,42,44,45 Ranges of mean utilities were 0.67 to 1, 0.72 to 1, and 0.82 to 1 for untreated BCC (or where treatment was unspecified), BCC treated with physical treatments, and BCC treated with other treatments, respectively. The utility for suspected but unconfirmed BCC was 0.97 (SD = 0.04). 41 For advanced BCC, Shingler et al 44 found utility values ranged from 0.67 (SD = 0.25) for progressed disease with 6 cm growth to 0.94 (SD = 0.08) after complete treatment response.

| Utilities pre-and post-treatment
Five studies compared pre-and post-treatment health states, with generally higher utility values posttreatment. 32,33,36,39,45 Treatment subgroups, where specified, included surgical excision (simple excision, flap excision, and Mohs surgery), cryotherapy, ingenol mebutate, electrochemotherapy and bleomycin, radiation therapy, and nodal dissection. Serial measurements in three studies 32,33,36 showed an initial decrease in utility following initiation of treatment, which increased over the post-treatment period and reached or exceeded the baseline utility value. Two of these studies 32,33 measured repeated utility values at defined time periods after treatment.
Overall, untreated stage III and stage IV SCC and untreated advanced BCC produced the lowest utility values, whereas utilities obtained in patients with AK were significantly higher.

| EuroQol 5-dimension instrument
Nine studies used the EQ-5D questionnaire to measure health utility, which ranged from means of 0.45 to 0.96. [32][33][34][35][36][37][38][39][40] Five of these studies were conducted among patients with undifferentiated NMSC without further subgroup analysis. 32,34,37,38,40 One study reported on undifferentiated NMSC as well as separate subgroup analysis of AK, BCC, and SCC. 35 Two studies reported on AK, 33,39 and one focussed on SCC. 36 Utilities obtained for pre-treatment AK subgroups ranged from 0.88 to 0.93. 33,35,39 Utilities for undifferentiated NMSC (untreated or treatment unspecified) ranged from 0.72 to 0.88. 32 41 We did not meta-analyse the primary data since there was considerable heterogeneity regarding stage diagnosed, time since treatment, treatment type, and utility instrument or method. It is likely that anatomic location of the cancer influences the quality of life, 53,54 however, only a few studies reported this. The context of each study, such as the population demographics 55 and anatomic location, is an important consideration when selecting which utility estimates to use, so caution should be used when pooling utility values for economic analyses. 56 The method or instrument also influenced the utility weights reported. Higher values were found in studies using the standard gamble method, which could be attributed to the endowment effect 57 -a tendency for higher utility values due to the healthy general population's aversion to succumbing to illness, limited understanding of the natural history, or the influence of the presented clinical scenarios to the outcomes. 45 A similar effect was also noted using the time trade-off method. 44 The asymmetry of information and difference in the understanding of the disease process between health professionals and patients highlights the importance of piloting the feasibility of the valuation approaches, and the challenges with establishing a single value to a temporary and changing health state. 58 The EQ 5D 5L is the 5-level version of the EQ 5D 3L, the 3-level questionnaire. 70 SO ET AL. Only one study 35 measured the marginal utility of increasing cancer burden on patients, which compared the mean utility between participants with either BCC (0.87) or SCC (0.84) against participants with both BCC and SCC (0.80), a difference of 0.07 and 0.04, respectively. Similarly the mean utility for participants with a single diagnosis of AK (0.89) and SCC (0.90) was compared to participants with both AK and SCC (0.82), a difference of 0.07 and 0.08, respectively. 35 Since many patients develop multiple tumours and require ongoing monitoring and treatment, the experience of skin cancer is akin to a chronic disease and is not a temporary minor ailment as commonly portrayed. 35,59 Participants were drawn from countries with predominantly fair-skin populations, however there were no studies from Australia, which has the highest rates of skin cancer in the world. 1 The New Zealand studies 34,47 had small sample sizes and only studied health utilities for specific treatment modalities related to treatment of large skin cancers. Some of the excluded studies 60,61 used pooled utilities derived from other studies and applied the value to populations and countries distinct from the original sample population. Generalizability of utility values between different populations can be adequately assessed when there is demographic data and clearly defined health states based on clinical characteristics.
Six studies 34,43-47 had a sample size of under 100. In particular, Pil et al 36 used assigned baseline utility values and extrapolation of utilities for SCC stages II, III, and IV (diagnosed) due to insufficient sample size.
The two national population studies 37,38 had large subgroups of over 400 participants with NMSC, however the pooling of data limits further disease-specific analysis. Philipp-Dormston et al's study 35 recruited 1184 participants from hospitals and local medical practices in Germany and was assessed as low risk of bias, therefore, based on these factors, could be considered to have produced reliable utility values.

| Limitations
Studies that reported mapped utilities or used instruments mapped by an algorithm to EQ-5D were excluded. For health economic modelling purposes, it is preferable to use utility values obtained directly from instruments and to reserve mapped values for when such data is not available. 62 Intrinsically the quality of mapped values relies on the accuracy of mapping algorithms. Also, non-utility based measures such as the DLQI may be more suitable for chronic, benign skin conditions such as eczema than NMSC as the instrument is sensitive to changes in level of discomfort related to itchiness and irritation, but not to treatment-related scarring or disfigurement nor to patient anxiety about recurrence. [63][64][65][66][67][68] However, instruments that have not yet been mapped to utility values, such as Skindex-16, 17 the Skin Cancer Index, 69 and BaSQoL, which are NMSC-specific and include measures of sun protective behaviours after skin cancer diagnosis and worries about treatment, 18 may also yield further insights into quality of life when compared to generic tools, such as EQ-5D, where subtleties of the quality of life experience may be lost. 35 One of the key advantages of generic tools is the transferability of results to health economics studies. Hence, there may be value in conducting further research in maximizing the sensitivity of quality of life data that are used in health economics studies. Studies in languages other than English were excluded and although this would represent few studies, it is possible that NMSC utilities differ by race. Authors of 12 short-listed conference abstracts were unable to provide further information, so only limited information could be extracted for these. Overall, this review highlights the paucity of evidence in the literature, with only seven studies 35,36,41,42,44,45,47 that reported utility values specifically for patients with BCC or SCC, suggesting further research is needed to obtain accurate and reliable utility values. Future primary research should aim for larger sample sizes with a priori specification of the required sample sizes for assessing utilities to a prespecified level of precision and report utilities for different subgroups defined by age, anatomic location, treatment status and the period of time post-treatment due to the impact of disfigurement and discomfort. 53 Patient comorbidity may also have an impact on the quality of life, so baseline clinical information would be useful information. Early detection of skin tumours leads to improved outcomes in most cases so future research should aim to demonstrate the likely higher posttreatment utility values where skin tumours have been treated early.
In conclusion, this systematic review found that most health utilities clustered around relatively high mean values of 0.8 to 1 for NMSCs and their precursors, with lower utility values for more advanced cancers. Although there are considerable difficulties comparing values obtained from studies using different methods, this seems to indicate small decrements to quality of life associated with the clinical management of most NMSC and AK. There was substantial variability in the utilities for different skin cancer subtypes, treatments and methods. Developing an accurate and comprehensive catalogue of NMSC health utilities in different populations is crucial for future health economic evaluations to adequately inform policy and resource allocation decisions related to skin cancer early detection and management.