An economic evaluation of the randomized controlled trial of topical corticosteroid and home‐based narrowband ultraviolet B for active and limited vitiligo (the HI‐Light Vitiligo Trial) *

Economic evidence for vitiligo treatments is absent.


Summary
Background Economic evidence for vitiligo treatments is absent. Objectives To determine the cost-effectiveness of (i) handheld narrowband ultraviolet B (NB-UVB) and (ii) a combination of topical corticosteroid (TCS) and NB-UVB compared with TCS alone for localized vitiligo. Methods Cost-effectiveness analysis alongside a pragmatic, three-arm, placebo-controlled randomized controlled trial with 9 months' treatment. In total 517 adults and children (aged ≥ 5 years) with active vitiligo affecting < 10% of skin were recruited from secondary care and the community and were randomized 1: 1: 1 to receive TCS, NB-UVB or both. Cost per successful treatment (measured on the Vitiligo Noticeability Scale) was estimated. Secondary cost-utility analyses measured quality-adjusted life-years using the EuroQol 5 Dimensions 5 Levels for those aged ≥ 11 years and the Child Health Utility 9D for those aged 5 to < 18 years. The trial was registered with number ISRCTN17160087 on 8 January 2015. economic results. A full and detailed trial report will be published within the National Institute for Health Research journal and copyright retained by the Crown.

Conflicts of interest
All authors have completed the ICMJE uniform disclosure form at http://www.icmje.org/conflicts-of-interest. All authors' organizations received financial support from the trial funder in order to deliver the submitted work; no authors received any additional support from any organization for the submitted work; no authors reported financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years; no authors reported other relationships or activities that could appear to have influenced the submitted work. H.C.W. is director of the NIHR Health Technology Assessment Programme but had no role in the funder's review of the research. J.R.I. is Editor-in-Chief of the BJD but had no role in the publisher's review of the submitted work.
*Plain language summary available online DOI 10.1111/bjd.19554 Results The mean AE SD cost per participant was £775 AE 83Á7 for NB-UVB, £813 AE 111.4 for combination treatment and £600 AE 96Á2 for TCS. In analyses adjusted for age and target patch location, the incremental difference in cost for combination treatment compared with TCS was £211 (95% confidence interval 188-235), corresponding to a risk difference of 10Á9% (number needed to treat = 9). The incremental cost was £1932 per successful treatment. The incremental difference in cost for NB-UVB compared with TCS was £173 (95% confidence interval 151-196), with a risk difference of 5Á2% (number needed to treat = 19). The incremental cost was £3336 per successful treatment. Conclusions Combination treatment, compared with TCS alone, has a lower incremental cost per additional successful treatment than NB-UVB only. Combination treatment would be considered cost-effective if decision makers are willing to pay £1932 per additional treatment success.
What is already known about this topic?
• Vitiligo is a common skin condition with significant psychological impact. • Topical corticosteroids (TCS) are standard care for vitiligo. Narrowband ultraviolet B (NB-UVB) is only available in secondary care as full-body treatment.
• Economic evidence for handheld NB-UVB in combination with TCS is absent.
What does this study add?
• Combination treatment, compared with TCS alone, has the lowest incremental cost per successful treatment. Whether this is considered cost-effective depends on decision makers' judgement on how much they are willing to pay to achieve a successful treatment.
• Generic utility instruments, such as the EuroQol 5 Dimensions 5 Levels, may not be appropriate for vitiligo studies due to high ceiling effects. Measurement of quality of life for this condition warrants further research.
• This study provides results that can be compared with those of new emerging vitiligo treatments.
A 2018 systematic review showed that the economic evidence for vitiligo treatment is virtually nonexistent. 1 One of two studies identified in this review estimated the annual direct cost of treating vitiligo in the USA to be $15 million for the price year 2004Á 2 The other study demonstrated that 32Á5% of people with vitiligo would be willing to make a one-off payment of €5000 for a cure (2006 price year), 3 allowing an estimate of the maximum potential for benefit should a 'cure' be found. Although these papers indicate the cost to an affected person and healthcare system, they do not provide evidence to inform resource allocation decisions. No papers were identified that undertook full economic evaluations (those that compare costs and benefits of two or more interventions) 4 of vitiligo treatments alongside clinical trials or as economic modelling. This paper reports the first full economic evaluation of treatment for localized, nonsegmental vitiligo, including the current standard treatment topical corticosteroids (TCS) and new treatment [home-based narrowband ultraviolet B (NB-UVB)], alone and in combination with TCS, with the aim of estimating the cost-effectiveness of these treatments for the UK National Health Service (NHS). Additional explanations of the terms used in this paper are provided in Appendix S1 (see Supporting Information).

Materials and methods
This health economic evaluation estimated the within-trial cost-effectiveness of (i) active handheld NB-UVB compared with TCS (standard care) and (ii) the combination of active handheld NB-UVB plus TCS compared with TCS (standard care). Estimates were made in terms of cost per additional treatment success (henceforth referred to as treatment success) at the end of the treatment period (9 months) for the treatment of limited, nonsegmental vitiligo, using individual-level data collected within the trial. A treatment period of 9 months was chosen to reflect clinical practice, where clinical experience and clinical guidelines suggest that treatment should be initiated for a minimum of 3-4 months, but that treatment would normally be required for a longer period in order to achieve a clinically meaningful treatment response.
A secondary objective was to undertake cost-utility analyses for those aged ≥ 11 years using the EuroQol 5 Dimensions 5 Levels (EQ-5D-5L) and separately for participants aged < 18 years using the Child Health Utility 9D (CHU-9D). Typically, a cost-utility analysis would form the primary analysis as it enables decision makers to compare the cost-effectiveness of a range of interventions for different conditions on a common scale. As utility is measured differently in adults and children, a common cost-utility analysis was not possible, so a clinical outcome was used. Also, cost-utility instruments are considered less effective at capturing the psychological impact on quality of life, which is considered to be more important than physical impacts in vitiligo. A priori we were also sceptical that available generic utility instruments would capture the health-related quality-of-life aspects that people living with vitiligo experience.
The evaluation was undertaken in line with published guidelines for the economic evaluation of healthcare interventions. 4-8 A health economics analysis plan was written and approved before the trial database was locked. A full trial report will be available through the NIHR journal series, 9 and the clinical results paper is available in this journal. 10 The trial was conducted in the UK NHS, which provides publicly funded healthcare that is largely free of charge at the point of use. Therefore, the analysis was primarily from an NHS perspective, in keeping with the National Institute for Health and Care Excellence reference case. 8 In a sensitivity analysis, out-of-pocket costs incurred by participants (or parents or guardians) are presented reflecting a personal perspective.

Resources use and costs
The primary analysis captured the intervention costs (including any side-effect costs) to the NHS and the participant's wider use of the NHS (including primary care visits; secondary care outpatient, inpatient and accident and emergency visits; and prescriptions) as a result of vitiligo. Participants' personal out-of-pocket expenses (for example, camouflage or makeup, suncream and sun care) incurred from vitiligo were also captured in a separate sensitivity analysis taking a broader perspective. Participant time burden for home treatment was not costed, but is reported elsewhere. 9,10 Resource use data for the intervention phase were collected at 3, 6 and 9 months using information recorded by participants in daily diaries and collated by the researcher at followup visits. Resource use related to the intervention and side-effects was recorded in clinical reports forms. Further questionnaires collected resource use data at 12, 15, 18 and 21 months for the follow-up phase.
Intervention cost was estimated at the individual level. Participants randomized to NB-UVB alone were also given a placebo ointment while those in the TCS alone group received a dummy NB-UVB device. The dummy devices and placebo ointment were not costed.

Narrowband ultraviolet B device
The cost of the handheld device was estimated using the manufacturer's purchase price divided by an annuity factor (interest rate 3Á5%, 5 years) to give an equivalent annual cost. The equivalent annual cost was divided by 12 months and multiplied by 9 to reflect the 9-month timeframe. The purchase prices of personal protective equipment (goggles and glasses) were included at full cost as these are unlikely to be as durable as the devices. Costs of the quality-assurance process for the devices were included. Device repair and replacement costs were not included in the analysis. Faulty devices were replaced in the study, although in practice some might be repaired. Time spent by investigators training participants on using the device was recorded and costed.

Topical corticosteroid
Participants in the TCS intervention group were supplied with two 90-g tubes of mometasone furoate 0Á1% ointment (Elocon â 0Á1% Ointment; Merck Sharp & Dohme, Hertford, UK). TCS costs were sourced from the Prescription Cost Analysis for 2017 11 and had the National Average Discount Percentage of 7Á37% deducted. 12 The professional pharmacist fee of £1Á29 was added, assuming that a single tube would be prescribed at any one time. Additional ointment requested by participants was recorded and costed.
Trial participants in all treatment groups were offered appointments with a dermatologist at 0, 3, 6 and 9 months, as we assumed in the analysis that this would happen in routine care. These were costed even though they cancel each other out between treatment groups. Side-effects requiring medical attention from either treatment were recorded as one type of unscheduled contact. Unit costs were identified from published sources (Table 1) and were valued in UK pound sterling 2017. 13,14 Patient-reported estimates of out-of-pocket costs resulting from vitiligo were captured.

Clinical outcome: treatment success
The primary clinical outcome measure in the HI-Light trial was participant-reported treatment success, measured at 9 months, using the Vitiligo Noticeability Scale. 15 Treatment success, a binary outcome, was defined by whether the participant responded that their target vitiligo patch was 'a lot less noticeable' or 'no longer noticeable' in response to the question: 'Compared with the start of the study, how noticeable is the vitiligo now?'. Because no previous studies have compared the treatments or outcome used in this study, we used a single study-based estimate of effectiveness in the cost-effectiveness analysis.

Quality of life
Quality-adjusted life-years (QALYs) were estimated in secondary analyses using utility scores obtained from the CHU-9D in the analysis focused on children aged < 18 years, [16][17][18] and the EQ-5D-5L instrument for participants aged ≥ 11 years, 19 For participants aged 5-6 years, the CHU-9D was completed by parental proxy. For all other ages these instruments were self-completed. We chose to use just one version of the EQ-5D-5L in the study for consistency. We chose the CHU-9D for the youngest participants because the EQ-5D-Y (youth) does not currently have a UK valuation set.
Utility measurements were collected in clinic at baseline and 9 and 21 months to reflect the likely timeframe for observing a clinically meaningful treatment response and in order to observe whether any response found was sustained in the longer term. In the cost-utility analysis, quality-of-life instrument responses were converted to utility scores using the EQ-5D-5L crosswalk 20 UK preference weights in line with current recommendations. 21,22 The CHU-9D was valued using the UK value set. 16 Following this, the utility values were used to calculate QALYs generated over the trial treatment period of 9 months, using both linear interpolation and area-under-thecurve analysis with baseline adjustment. 23

Economic analysis
The economic primary analysis was performed on the full analysis set. In line with the primary statistical analysis, 10 multiple imputation was used to account for missing primary outcome data at 9 months. Cost analyses employed multiple imputation with chained equations using MI impute in Stata (StataCorp, College Station, TX, USA), generating 60 (m = 60) datasets using predictive mean matching and separately by treatment allocation as reported by Faria et al. 24 Given the 9month time horizon, costs and benefits were not discounted.
Mean AE SD resource use and cost per participant were estimated for each randomized group. The mean difference with 95% confidence interval (CI) in resource use and cost between arms (NB-UVB vs. TCS and combination treatment vs. with TCS) is presented. Costs and QALYs were adjusted for age and location of target patch, as well as baseline utility, using seemingly unrelated regression. 25 Nonparametric bootstrapping was used to determine sampling uncertainty surrounding the mean incremental cost-effectiveness ratios by generating 10 000 estimates of incremental costs and benefits. These estimates were used to produce cost-effectiveness acceptability curves to show the probability that each intervention arm is cost-effective at different values of willingness to pay.
Other than preplanned secondary analysis based on the different utility instruments used (EQ-5D-5L and CHU-9D), no subgroup analyses were undertaken. The secondary outcome for the economic evaluation was QALYs per participant over 9 months. The mean AE SD utility and mean AE SD QALYs per participant per randomized group were estimated, as was the mean difference (95% CI) in QALYs between arms (NB-UVB vs. TCS, and combination treatment vs. TCS) adjusted for age and location of target patch. In secondary analyses, the reported economic analysis used a cost-effectiveness threshold of £20 000 per QALY. 8 All analyses were conducted in Stata MP4 version 15.
Sensitivity analyses were undertaken to explore key uncertainties including (i) comparing multiple-imputation analysis to a complete-case analysis, (ii) varying NB-UVB device costs (zero and double the price in the primary analysis), (iii) a wider cost perspective including vitiligo out-of-pocket costs, (iv) limiting analysis to participants with good adherence (defined as > 75% adherence), and (v) extending the time horizon to 21 months to include the 12-month follow-up period.
It was expected that the majority of costs and benefits would be captured in the treatment period such that a priori it was not considered necessary to develop a decision-analytical model for a longer timeframe. This proved appropriate, as quality-of-life scores were similar between treatment arms at 21 months. 10

Data sharing
Anonymized patient-level data are available from Dr Jonathan Batchelor (jonathan.batchelor@nottingham.ac.uk) upon reasonable request.

Results
The baseline characteristics of the participants included in the cost-effectiveness analysis are described in Thomas et al. 10 With imputation, 517 participants were included (398 adults, 119 children; 173 TCS, 169 NB-UVB and 175 combined treatment).

Intervention costs
The mean drug and training costs and numbers of devices, goggles, glasses, dermatology appointments and unscheduled visits or telephone appointments by group are reported in Table 2 and the mean costs in Table 3. The mean cost of the intervention per participant was £583 AE 29Á6 for TCS (standard care), £753 AE 59Á2 for NB-UVB and £792 AE 94Á6 for combination treatment. Details of the time and cost of qualityassurance processes are shown in Table S1 (see Supporting Information).
The training time was a mean of 73Á1 min for NB-UVB and 69Á2 min for combination treatment, noting that all participants received both a device and ointment (dummy devices and placebo ointment were not costed).

Wider resource use and costs
Wider healthcare resource use (primary care, secondary care and medicines) for vitiligo beyond those required for the intervention were not significantly different between groups ( Table 2). Patients with vitiligo reported low NHS healthcare usage. Table 3 displays the mean costs per participant by treatment group using available case data. The overall mean cost per participant for NB-UVB was £775 AE 83Á7, compared with £600 AE 96Á2 TCSan unadjusted mean difference in cost of £175 (95% CI 153-197). Combination treatment had overall mean costs per participant of £813 AE 111; compared with TCS this gave an unadjusted mean difference of £213 (95% CI 188-238) per participant. These figures suggest that the costs of the interventions were not offset by reductions in wider healthcare resource use related to vitiligo.

Primary economic analysis
Cost-effectiveness analysis of narrowband ultraviolet B compared with topical corticosteroid (standard care) The adjusted incremental difference in cost was £173 (95% CI 151-196). The adjusted risk difference for NB-UVB compared with TCS was 5Á2%. This equates to a number needed to treat of 19; in other words, 19 participants would need to be treated for one of them to gain treatment success. The adjusted incremental cost was £3336 per additional successful treatment, which was estimated by dividing the adjusted incremental difference in cost, £173, by the adjusted risk difference, 0Á052. Figure 1(a) shows the probability that NB-UVB is cost-effective at different possible levels of willingness to pay for an additional treatment success; probability increases as willingness to pay increases. Figure 1(a) shows considerable uncertainty surrounding the decision as to whether NB-UVB, compared with TCS, represents value for money, as there is always at least 40% probability of making the wrong decision if choosing to fund NB-UVB alone below a threshold value of willingness to pay of £10 000 per additional treatment success.
Cost-effectiveness analysis of combination treatment compared with topical corticosteroid (standard care) The adjusted incremental difference in cost was £211 (95% CI 188-235). The adjusted risk difference for combination treatment compared with TCS was 10Á9%. This equates to a number needed to treat of 9. The adjusted incremental cost was £1932 per additional successful treatment. Figure 1(b) shows the probability that combination treatment is cost-effective at different possible levels of willingness to pay for an additional treatment success. It shows that combination treatment is likely to be cost-effective if decision makers are willing to pay more than £3000 per additional treatment success, as the probability of making the wrong decision is < 50%.
Sensitivity analyses exploring key uncertainties in the economic evaluation are summarized in Table S2 (see Supporting Information). Limiting analysis to only adherent participants made the most difference to the incremental cost-effectiveness ratio (£1151 for combination treatment compared with TCS, and £1394 for NB-UVB compared with TCS). Those who were adherent to treatment were more likely to be cost-effective to treat.

Secondary economic analysis
In total 248 (55%) trial participants reported having no problems on any of the five domains of the EQ-5D-5L at baseline, suggesting that over half of the sample started the study in perfect health as defined by EQ-5D-5L. To put this value into perspective, in a general population sample from England the number of participants reporting no limitations on any dimension of the EQ-5D-5L was 43Á9%. 26 Thus, the ceiling effect in this study can be considered large and of an order such as to limit the discriminatory power of the instrument for this patient population. Similar levels of ceiling effect were observed at subsequent follow-up. Similarly, for the CHU-9D, 30% of participants aged < 18 years had no problems according to any of the nine dimensions on the CHU-9D at baseline. Anxiety and depression on the EQ-5D-5L, and worry, tiredness and sleeping on the CHU-9D were the domains for which problems were reported most commonly. No floor effect was observed at any timepoint on either instrument.
As these high ceiling ratios suggest that these instruments are unlikely to be able to detect change, we report the Medication 0Á12 AE 0Á50 (138) 0Á08 AE 0Á35 (133) À0Á04 (À0Á14 to 0Á06) 0Á09 AE 0Á34 (141) À0Á03 (À0Á13 to 0Á07) Out-of-pocket purchases 0Á40 AE 1Á44 (141) 0Á28 AE 0Á88 (137) À0Á12 (À0Á40 to 0Á16) 0Á31 AE 1Á27 (144) À0Á09 (À0Á41 to 0Á23) CI, confidence interval; NB-UVB, narrowband ultraviolet B; TCS, topical corticosteroid. a Includes the number of NB-UVB devices only. b Participants could choose to have more than one set, for instance if they needed a parent or partner to help them deliver the treatment.
mean utility estimates in Tables S3 and S4 and the costutility analyses in Table S5 (see Supporting Information). With this limitation in mind, both NB-UVB and combination treatment compared with TCS (standard care) had cost-utility ratios within accepted thresholds (< £20 000 per QALY) for the sample aged ≥ 11 years (NB-UVB was more superior to TCS than combination treatment was vs. TCS, in contrast to the cost-effectiveness analysis). Neither treatment was cost-effective in the analyses of those participants aged < 18 years, but this may reflect the small sample size (n = 119).

Discussion
We present the first full economic evaluation of treatments for vitiligo using the standard care TCS as the comparator. The additional cost of the combination treatment was not offset by NHS cost savings but did result in significant treatment success over the 9-month treatment period. This benefit could be gained if decision makers were willing to pay more than the adjusted incremental cost of £1932 per additional successful treatment. NB-UVB was less costly than combination treatment but also less effective, such that the incremental cost per successful treatment was higher than for combination treatment, suggesting that the NHS would get better value for money from combination treatment than NB-UVB therapy alone. There is currently no evidence to indicate how much a decision maker would be willing to pay for an additional treatment success as defined in this study. Should the decision makers' willingness to pay per additional treatment success be low, then uncertainty surrounding the decision to fund combination treatment is high. Treatment options are limited for vitiligo and existing treatments are used little in the NHS, which may be due to treatments not being offered rather than absence of need. 27 Cost-effectiveness analysis was undertaken as the primary analysis because it enabled us to analyse all participants together, irrespective of age. We had a prior belief that generic utility instruments may not fully capture the health-related quality-of-life impairment of people living with vitiligo. This was supported by high ceiling effects on the EQ-5D-5L and CHU-9D at baseline, such that there was no capacity to measure any gain using these instruments for many participants. The cost-utility analysis gave different results from the clinical and cost-effectiveness results, in that NB-UVB appeared more cost-effective than combination treatment, compared with TCS, for those aged ≥ 11 years.
There was also a difference in results between the cost-utility analyses undertaken by age; the new interventions were estimated as cost-effective in those aged ≥ 11 years but not in those aged < 18 years. This could reflect the different utility instrument used, but more likely reflects the small sample size of the group aged < 18 years and the fact that there was a lot of uncertainty around the QALYs gained, as the gain between groups was very close to zero in all comparisons. Therefore, more weight should be attached to the clinical effectiveness results and further work to explore the validity of the EQ-5D-5L and CHU-9D in this patient group is warranted, given the high ceiling effect observed in this study. It may be that a disease-specific utility instrument needs to be developed for vitiligo.
Sensitivity analyses suggested that a wider perspective, the cost of the NB-UVB device, and a method of dealing with missing data did not change the conclusions reached. Incremental cost per treatment success was lowest for those with greatest adherence.
New treatments such as Janus kinase inhibitors are being developed for vitiligo and are likely to be costly. The relatively low cost of the interventions assessed in this trial may make them affordable when resources are limited. The trial has yielded useful cost-effectiveness data, which can be used for future comparisons with novel treatments.
A strength of the study was that the HI-Light trial was a large, pragmatic trial of home interventions for people with active, limited vitiligo that controlled for common causes of bias. Retention throughout the trial was challenging, and the treatments placed considerable time burden on participants. Because < 50% responded to secondary outcomes at 21 months, a longer-term economic evaluation to 21 months was not undertaken, which is a limitation of the present study. However, given that treatment effects beyond the 9month period were not sustained one can assume that the cost-effectiveness of the interventions would likely decline over time if treatments were not continued.
In conclusion, combination treatment compared with TCS alone has a lower incremental cost per successful treatment than NB-UVB vs. TCS, but whether this is considered cost-effective will depend on how much healthcare decision makers are willing to pay to achieve a successful treatment. The fact that vitiligo has few treatment options available, and the likely high cost of newer treatments being developed, may influence these decisions.
for help in trial design and identifying recruiting sites. The UK DCTN is grateful to the British Association of Dermatologists and the University of Nottingham for financial support of the Network. We are grateful to Dermfix Ltd (www.dermf ix.uk) for providing the NB-UVB devices and dummy devices at a reduced cost. Dermfix Ltd had no role in the design, conduct or analysis of the trial. Members of the Centre of Evidence Based Dermatology's Patient Panel provided input into the trial design and conduct, and dissemination of trial findings. Thanks also to the people who responded to online surveys to inform the trial design. We thank Emma McManus for review and comments on the Health Economics Analysis Plan. A full list of contributors is provided in Appendix S2 (see Supporting Information).

Supporting Information
Additional Supporting Information may be found in the online version of this article at the publisher's website: Appendix S1 Supplementary index of definitions. Table S1 Quality-assurance process (time and costs) for narrowband ultraviolet B devices.