FormalPara Key Points for Decision Makers

Tirbanibulin 1% ointment is estimated to be cost saving and produce similar outcomes at one year compared with the three most frequently prescribed treatments in Scotland.

Tirbanibulin is associated with a lower probability of severe local skin reactions, and has a shorter treatment duration, than alternative AK treatments - this may improve treatment adherence.

1 Introduction

Actinic keratosis (AK) is a chronic skin condition caused by long-term exposure to sunlight. Ultraviolet (UV) radiation from the sun induces abnormal keratinocyte proliferation, resulting in damage to the skin that presents as scaly red lesions [1]. AK is common in adults with fair skin and a history of sun exposure. It is estimated that 19–24% of people over the age of 60 years old in the UK have at least one AK lesion, most frequently on the skin of the head or hands [2].

Symptoms experienced by people with severe AK can include pain, itchiness, and bleeding [3]. AK may develop into squamous cell carcinoma (SCC) if left untreated. Although individual lesions carry a very low risk of malignancy, the overall risk increases with the number of lesions present [2]. A person with an average of 7.7 lesions has approximately a one in ten chance of developing SCC within ten years [3].

Management of AK in Scotland (and the rest of the UK) includes the use of emollients and sun protection in the first instance. Several active topical treatments applied as a cream or gel are available, and photodynamic therapy and cryotherapy may also be used to target individual lesions and/or the surrounding field [2]. Many of the current treatments for AK are associated with significant rates of adverse events (primarily local skin reactions), which need to be treated with steroids, and can require relatively long courses of treatment (e.g., 28 days for imiquimod cream) [1, 4]. The license for one treatment, ingenol mebutate gel, was recently suspended due to treatment-related adverse events [5].

Tirbanibulin is an ointment formulation designed to treat AK on the face or scalp. It is a non-invasive, short-term (five consecutive days) treatment. Two phase III clinical studies have reported statistically significant improvements in complete clearance of AK lesions 57 days after the initiation of treatment with tirbanibulin when compared with placebo (vehicle ointment without active ingredient) [6, 7].

An economic model was developed to estimate the cost-effectiveness of tirbanibulin for the treatment of AK on the face or scalp from the perspective of the Scottish Healthcare System. The model was developed for submission to the Scottish Medicines Consortium (SMC), which requests evidence of clinical and cost-effectiveness for all new therapies. In December 2021, the SMC issued advice to NHS Scotland Boards recommending the use of tirbanibulin [8].

2 Methods

2.1 Model Overview

A decision tree model was developed in Microsoft Excel (Microsoft Corporation, Redmond, WA, USA) to assess the cost-effectiveness of tirbanibulin compared with three stand-alone comparators identified as being most relevant to the Scottish setting by clinical experts (described further in Sect. 2.5), i.e. diclofenac sodium gel 3%, imiquimod cream 5%, and fluorouracil cream 5%. Imiquimod cream 3.75%, fluorouracil (5 mg) with salicylic acid (100 mg) and photodynamic therapy were identified as having a small market share in Scotland and were, therefore, included in scenario analyses only (Sect. 2.4.5). Clinical experts advised that cryotherapy was used relatively less frequently, and exclusively in a hospital setting for a subgroup of patients, in Scotland (Sect. 2.5). Therefore, these experts advised that cryotherapy would not be an appropriate comparator.

Tirbanibulin 1% ointment was included in the analysis as per its licensed indication for adults with AK of the face or scalp (one single-use sachet applied to the treatment area once daily for five days) [9]. All comparator treatments were used at the recommended dosages and durations (described further in Sect. 2.2).

Figure 1 shows the decision tree framework used to estimate the length of time an average patient spent with one or more AK lesions over the one-year time horizon of the model. A one-year time horizon was considered sufficient to capture the costs and outcomes following one round of treatment. Due to the limited availability of effectiveness data for patients with prior exposure to treatment, it was assumed patients would not be retreated if treatment was ‘unsuccessful’.

Fig. 1
figure 1

Base-case model structure

Full size image

Treatment was ‘successful’ if all clinically visible lesions in the treatment area had cleared by the end of the time horizon, and ‘unsuccessful’ if any treated lesions remained. Due to the short time horizon, and a lack of data describing the efficacy of repeated rounds of treatment, it was assumed that all patients ceased treatment at the end of the recommended treatment duration and did not undergo further rounds if the initial treatment was ‘unsuccessful’.

In the base-case analysis, it was assumed that patients in whom treatment was ‘successful’ would remain lesion-free for the remainder of the time horizon (i.e., that no new lesions would develop). Similarly, it was assumed that patients in whom treatment was ‘unsuccessful’ would have clinically visible lesions for the entire time horizon (i.e., it was not possible for lesions to regress naturally). Both assumptions were made due to a paucity of data in the literature reporting the frequency of new lesion development and lesion regression.

Patients in both branches of the decision tree could experience adverse events (severe local skin reactions) associated with treatment.

The model was built from the perspective of the Scottish Healthcare System. Quality-adjusted life-years (QALYs) and costs associated with each branch of the decision tree were aggregated for each cohort (tirbanibulin vs. comparator) and compared with estimated relative cost effectiveness. Societal costs and benefits were not included within the analysis, in alignment with SMC guidelines. Discounting of costs and benefits was not necessary in the base-case analysis because the time horizon was one year.

2.2 Data Sources and Model Parameters

2.2.1 Patient Population

The model cohort was based on the population of the two phase III trials of tirbanibulin (NCT03285477 and NCT03285490), which aligned with the marketing authorisation for tirbanibulin [10]. All patients were assumed to have a diagnosis of clinically typical, visible, and discrete AK lesions on the face or scalp. Patients were required to have a contiguous treatment area measuring 25 cm2 and containing four to eight lesions. Based on patient characteristics from the phase III trials, the population entering the model was 87% male, with a mean age of 70 years. Further information regarding the model cohort is provided by Blauvelt et al. [10].

2.2.2 Clinical Effectiveness

The probability of complete clearance (reduction of 100% in the number of AK visible lesions in the application area) with tirbanibulin was estimated based on data from the phase III trials [11]. Across the two trials, a total of 173/353 patients treated with tirbanibulin experienced complete clearance, while 180 patients treated with tirbanibulin did not experience complete clearance, equating to an odds of complete clearance of 0.962 [11]. By comparison, only a total of 30 out of a total 349 patients (across both trials) treated with the vehicle placebo experienced complete clearance.

A network meta-analysis (NMA) of relevant randomised controlled trials (RCTs) was performed to determine the odds ratio (OR) of complete clearance associated with each of the comparators compared to tirbanibulin. Complete clearance is a standard measure used to inform the efficacy of all comparators in the analyses. The RCTs included within the NMA were identified from a systematic review conducted in 2020 to assess the comparative efficacy and safety of tirbanibulin and comparators in the treatment of AK. Further summary details on the systematic review and NMA are provided in electronic supplementary material (ESM) 1, with full details as reported by Heppt et al. [12].

The relative probability of complete clearance associated with each comparator was estimated by applying ORs from the NMA to the odds of complete clearance for tirbanibulin. The odds were then converted to probabilities using the following formula: \(Probability= \frac{Odds}{1+Odds}.\)

With the exception of diclofenac sodium gel and photodynamic therapy, there was no statistically significant difference in the probability of complete clearance for any of the comparators (including the additional comparators used in the scenario analyses) when compared with tirbanibulin (Table 1). In the base-case analysis, for comparisons where there was no statistically significant difference in treatment effectiveness, it was assumed that the probability of treatment success was equivalent to tirbanibulin.

Table 1 Clinical input data used in model
Full size table

It was assumed that complete clearance of lesions occurred at the time to maximum treatment effect reported in the literature (Table 1). Therefore, patients receiving treatments with a shorter time to maximum treatment effect spent less time with lesions present than those receiving treatments with a longer time to maximum treatment effect. This meant that the disutility associated with AK was applied to such patients over a shorter period.

2.2.3 Adverse Events

The model also captured the costs and disutilities incurred by patients if they experienced a severe local skin reaction during, or following, treatment. The proportion of patients experiencing a severe local skin reaction differed depending on the treatment received, and was determined from a targeted literature review (Table 1). It was not possible to estimate the difference in frequency of severe local skin reactions between tirbanibulin and each comparator through an NMA due to differences in the definitions reported and insufficient/inconsistent incidence data.

2.2.4 Costs

Direct medical costs associated with treatments and treatment-related adverse events were captured in the model. The frequency of each resource use was multiplied by the unit cost (Table 2). Costs were aggregated to produce the total costs of treatment with tirbanibulin and each comparator. Costs were sourced from recognised national sources wherever possible. All unit costs were based on the 2019/2020 price year.

Table 2 Costs and resource use data used in the model
Full size table

The number of tubes of diclofenac sodium 3%, fluorouracil 5% and fluorouracil (5 mg) with salicylic acid (100 mg) required per treatment course were associated with uncertainty because the treatment licenses do not specify how many grams should be used per application. It was assumed that patients used two 50 mg tubes of diclofenac sodium 3% per treatment course because the studies informing the NMA reported that patients used 0.5 mg twice daily for up to 90 days and it was accepted that some treatment may be wasted. The studies used to inform the efficacy of fluorouracil 5% and fluorouracil (5 mg) with salicylic acid (100 mg) did not report the number of tubes required per treatment course. The license for fluorouracil 5% notes that patients should use the cream once or twice daily for 3–4 weeks. Given this cream is available in a 40g tube, and the treatment duration is significantly less than diclofenac sodium 3%, it was assumed that patients would only require one tube of this treatment. Finally, given that fluorouracil (5 mg) with salicylic acid (100 mg) should be used daily for 12 weeks, it was assumed patients would incur two 25 mL units per treatment course. All assumptions regarding the number of tubes required were validated at the expert panel meeting.

All patients attended an appointment with a general practitioner (GP) upon entry to the model, in line with a previous SMC assessment for the treatment of AK [24]. A proportion of patients were also referred to a secondary care consultation with a dermatologist before receiving each treatment (Table 2). These proportions were treatment dependent and informed by an advisory board (described further in Sect. 2.5). These proportions differed by treatment because clinicians indicated that GPs would feel more comfortable prescribing certain treatments without referral than others (i.e., GPs would feel less comfortable prescribing a treatment that was associated with higher rates of severe local skin reactions). If treatment was ‘unsuccessful’, patients made an additional visit to a healthcare professional to determine the next steps in their therapy. It was assumed that a patient would return to the dermatologist if this was where they were initially referred, while the remainder of patients would return to their GP. If treatment was ‘successful’, it was assumed no further appointments with a healthcare professional were necessary. This assumption was varied in a scenario analysis. All patients experiencing an adverse event (i.e., a severe local skin reaction) incurred the cost of one GP appointment, regardless of the initial treatment received.

2.2.5 Health-Related Quality of Life

Utility data for AK were identified through a systematic review of the literature (further details are reported in ESM 1). A mean utility associated with AK was taken from a catalogue of dermatitis utilities [30] which reported a figure of 0.981. This utility value was informed from the completion of a time trade-off questionnaire (via interview) by nine patients diagnosed with AK upon visiting one of three US hospitals. These were the only relevant data identified, indicating a lack of robust utility data relating to AK.

A disutility of 0.019 (1–0.981) associated with AK was calculated using this source. This disutility was applied while any clinically visible AK lesions were present by subtracting 0.019 from the UK population norm utility value (adjusted for the age and sex profile of the modelled patient population) [31]. Patients were assigned the UK population norm utility value with no disutility applied if complete clearance of all AK lesions was achieved.

An additional disutility of 0.085 was applied to patients who experienced a severe local skin reaction. This disutility was also based on the completion of a time trade-off questionnaire by nine patients diagnosed with pruritus and related conditions upon visiting one of three US hospitals [30]. This value may represent an underestimation of the type of severe local skin reaction associated with AK treatment, which can include symptoms such as severe blistering of the skin (this point was raised by Scottish clinical experts during model validation). This disutility was applied for 8 weeks [32]. Therefore a further disutility of 0.013 was subtracted from the annual utility of patients experiencing a severe local skin reaction. Patients experiencing severe local reactions with imiquimod 5% and fluorouracil 5% are required to use topical steroids, which may cause further quality-of-life decrements [4]; however, these were not included in the model to be conservative against tirbanibulin.

2.3 Sensitivity Analysis

Deterministic sensitivity analysis (DSA) was conducted to account for first-order uncertainty around all clinical effectiveness, cost, health-related quality of life (HRQoL), and adverse event parameters. Parameters were varied using associated 95% confidence intervals where available. To account for the large uncertainty, parameters associated with healthcare professional resource use were varied by a range of 50%; all other parameters varied by ± 15%.

Probabilistic sensitivity analysis (PSA) was undertaken with 1000 model simulations. Beta distributions were fitted to the proportions of patients experiencing treatment success and adverse events. Gamma distributions were fitted to the disutility and resource use parameters. In the absence of data on the variability around the sampling distribution of mean values, the standard error was assumed to be equal to 15% of the mean.

2.4 Scenario Analyses

2.4.1 Recurrence

This scenario modelled the risk that patients in whom treatment was originally ‘successful’ could experience a recurrence of AK lesions (assumed to occur one year after treatment initiation). The analysis adopted a modified model structure, where the ‘successful’ branch included additional nodes for recurrence and non-recurrence (Fig. 2).

Fig. 2
figure 2

Schematic of the model structure in the recurrence scenario analysis

Full size image

Due to a lack of data available to perform an NMA on the probability of recurrence, all recurrence inputs were identified through a targeted literature review. The proportion of patients experiencing recurrence following complete clearance was dependent on the treatment received (Table 3). The cost of an additional visit to a healthcare professional was applied for all patients experiencing recurrence.

Table 3 Probabilities of recurrence within 1 year of treatment initiation associated with each treatment used in the recurrence scenario analysis
Full size table

Due to the short time horizon of the model, and the limited availability of effectiveness data for patients with prior exposure to treatment, it was assumed that patients would not be retreated if they experienced recurrence. Therefore, it was assumed that patients could not experience more than one recurrence event and no treatment costs were incurred by patients experiencing recurrence.

The time horizon of the analysis was extended to two years the impact of recurrence on HRQoL could be captured. Patients who experienced a recurrence were assumed to have clinically visible AK lesions (and the associated disutility was applied) for the remainder of the model time horizon. A discount rate of 3.5% was applied to QALYs after one year, in line with current SMC and Treasury guidance [35]. It was not necessary to apply the discount rate to costs because these were experienced upfront upon recurrence.

2.4.2 Point Estimates of Clinical Effectiveness

In this scenario, rather than assuming equivalence in clinical effectiveness where no statistically significant difference between tirbanibulin and the comparator was observed (this was the case for imiquimod 5% and fluorouracil 5%), point estimates for the probability of treatment success (complete clearance) were used (Table 1). There was no change from the base case in the values used for diclofenac sodium 3% because there was a statistically significant difference in clinical effectiveness compared with tirbanibulin. Tornado diagrams using these point estimates are also presented within the ESM.

2.4.3 Equivalent Time to Maximum Treatment Effect

In this scenario, it was assumed that complete clearance occurred at the same time point (90 days) for all treatments (i.e., time to maximum treatment effect was equivalent). Ninety days was considered to be an appropriate mid-point across the time to maximum effect for tirbanibulin and all comparators.

2.4.4 Additional Visit to a Healthcare Professional

In this scenario, it was assumed that patients would always return to the healthcare professional that initially prescribed the treatment, regardless of whether the treatment was ‘successful’ or ‘unsuccessful’.

2.4.5 Additional Comparators

Tirbanibulin was compared with photodynamic therapy, imiquimod 3.75% and fluorouracil 5 mg with salicylic acid (100 mg) creams, which were identified by Scottish clinical experts as having a small market share (Sect. 2.5).

2.4.6 No Severe Local Skin Reactions

It was not possible to estimate the difference in severe local skin reaction frequency between tirbanibulin and each comparator through the NMA due to differences in the definitions reported in the literature and insufficient/inconsistent incidence data. Therefore, in this scenario, it was assumed that no patients experienced any severe local skin reactions.

2.4.7 No Secondary Care Referrals

The proportion of patients referred to a secondary care consultation before receiving each treatment was informed by clinicians during an advisory board but remains uncertain. This scenario assumed that no patients required referral to secondary care before receiving any treatment.

2.4.8 Alternative Adverse Event Data

The proportion of patients experiencing severe erythema within a randomised trial of four treatments (fluorouracil 5%, imiquimod 5%, photodynamic therapy and ingenol mebutate) was used to inform the proportion of patients experiencing severe local skin reactions within this scenario [36]. The probabilities used in this scenario were 71.0%, 56.4% and 66.9% for fluorouracil 5%, imiquimod 5% therapy and ingenol mebutate, respectively. The study did not report such data for photodynamic therapy. The adverse event rates for treatments not reported within this study were not amended.

2.4.9 One Tube of Diclofenac Sodium 3% Per Treatment Course

This scenario conservatively assumed that patients would only receive, and incur the costs of, one tube of diclofenac sodium 3% per treatment course.

2.5 Model Validation

The face validity of the model was examined by a panel of Scottish clinical experts as part of an advisory board process. The expert panel consisted of three consultant dermatologists, two Scottish directors of pharmacy, a former principal pharmacist for the SMC, and two SMC health economic experts. The model structure and inputs were validated by the experts to ensure the model adequately captured costs and outcomes that were reflective of Scottish clinical practice. The expert panel was run in alignment with the Association of the British Pharmaceutical Industry code of conduct [37]. The modelling approach was presented to the experts, who were asked specific questions throughout; there were no instances in which the experts did not agree.

3 Results

3.1 Base-Case Analysis

Results over a one-year time horizon show that tirbanibulin is cost saving and is associated with a positive net monetary benefit (NMB) when compared with all three comparators (Table 4). Disaggregated cost results indicate that there are substantial savings in patient management costs (visits to primary and secondary care) when tirbanibulin is compared with diclofenac sodium 3%, imiquimod 5% and fluorouracil 5%. Tirbanibulin is also associated with higher QALYs than each comparator, and therefore the intervention is dominant and is predicted to be cost-effective at a threshold of £20,000 per QALY; however, differences in QALYs are negligible.

Table 4 Base-case results (per patient)
Full size table

3.2 Sensitivity Analysis

The results of DSA indicate that varying model input parameters does not have an impact on the direction of results (i.e., NMB does not become negative for any of the comparators). Key drivers across all comparisons are the probability of complete clearance for both tirbanibulin and the comparator, as well as the additional resource use required if treatment is ‘unsuccessful’. The proportion of patients requiring treatment with secondary care in the comparator arm is also a driver of results when tirbanibulin is compared against imiquimod 5% and fluorouracil 5%. Tornado plots from DSA are shown in ESM 2.

The results of the PSA are similar to the base-case analysis, with tirbanibulin being less costly and generating slightly more QALYs than all the comparators (ESM 2).

3.3 Scenario Analyses

Table 5 shows a summary of the results of the scenario analysis (full results are presented in ESM 2).

Table 5 Results of scenario analysis (per patient)
Full size table

Tirbanibulin remains cost saving compared with all comparators when the recurrence of AK is included in the model (scenario 1) (Table 5). The incremental cost becomes smaller when tirbanibulin is compared with diclofenac sodium 3% because there is a higher probability of recurrence following complete clearance, within 1 year of starting treatment, associated with tirbanibulin than diclofenac sodium 3%. Although the probability of recurrence is greater for tirbanibulin than imiquimod 5%, there is minimal change in incremental cost from the base-case analysis because this is outweighed by a greater proportion of patients visiting secondary care. The cost savings associated with tirbanibulin compared with fluorouracil 5% are higher in this scenario than in the base-case because recurrence rates are similar between the two interventions and a greater proportion of patients visit secondary care healthcare professionals after experiencing a recurrence with fluorouracil 5% than tirbanibulin.

In scenarios 2, 3, and 4 (using point estimates of clinical effectiveness, equivalent time to maximum treatment effect, and including an additional visit to a healthcare professional, respectively), tirbanibulin remains cost saving with a small positive impact on quality of life.

When additional comparators are included (scenario 5), tirbanibulin is cost saving compared with imiquimod 3.75% and photodynamic therapy, but not cost saving when compared with fluorouracil 5 mg with salicylic acid (100 mg). This is because it was assumed that all patients received fluorouracil with salicylic acid from the GP, with no referral to secondary care. Although tirbanibulin is not cost saving in this scenario, it is estimated to be cost-effective at a cost-effectiveness threshold of £20,000 per QALY, with an ICER of £2,946 and a positive NMB of £69. Tirbanibulin remains cost-effective in this scenario because it has a shorter time to maximum treatment effect and a lower rate of severe local skin reactions than fluorouracil with salicylic acid.

Tirbanibulin also remains cost saving when no costs or disutilities are applied to patients experiencing severe local skin reactions and alternative probabilities are used to determine the probability of patients experiencing local skin reactions (scenarios 6 and 8). Tirbanibulin remains cost-effective when no secondary referrals are required for any treatments and when only one tube of diclofenac sodium 3% is required per treatment course (scenarios 7 and 9, respectively). Tirbanibulin is not cost-effective when compared with fluorouracil 5% within scenarios 10 and 11; however, both of these scenarios are conservative and are worst-case scenarios.

4 Discussion

The results of the cost-effectiveness analysis suggest that tirbanibulin is cost saving versus diclofenac sodium 3%, imiquimod 5%, and fluorouracil 5% for treating AK on the face or scalp.

Tirbanibulin is associated with a greater quality of life than diclofenac sodium 3% because it is associated with a higher probability of complete clearance after a shorter time period, and has a lower proportion of patients incurring severe local skin reactions. A course of treatment of tirbanibulin is also cheaper than diclofenac sodium 3% and therefore tirbanibulin remains cost effective in all scenarios when compared with diclofenac sodium 3%.

Despite higher treatment costs, tirbanibulin is cost-effective when compared with imiquimod 5% in all scenarios because tirbanibulin is associated with a smaller proportion of patients incurring severe local skin reactions and requiring secondary care referrals.

There is greater uncertainty associated with the cost-effectiveness of tirbanibulin compared with fluorouracil 5%. Tirbanibulin is cost-effective in the base-case analysis because the increased treatment costs are outweighed by savings incurred by fewer secondary care referrals and a smaller proportion of patients incurring severe local skin reactions. However, tirbanibulin was not cost-effective versus fluorouracil 5% in three conservative scenarios: (1) when the proportion of patients incurring severe local adverse events is assumed to be zero; (2) when the point estimates of complete clearance are used in combination with no patients being referred to secondary care; and (3) when the proportion of patients incurring severe local skin reactions and being referred to secondary care are assumed to be zero.

Although tirbanibulin improves both costs and QALYs compared with the other treatments in the base-case analysis, the increases in QALYs are marginal. Given this, as well as the poor quality of available utility data for AK (value based on data from nine people), the results of the model can be considered as a cost-minimisation analysis.

There is a high degree of certainty that tirbanibulin is cost-effective at a threshold of £20,000 per QALY in the Scottish NHS. The main drivers of the cost-effectiveness model are the probability of treatment success (complete clearance of AK lesions), the percentage of patients referred to secondary care before receiving treatment, and the number of secondary care visits associated with ‘unsuccessful’ treatment. No inputs cause the results to change to a large enough extent that tirbanibulin is no longer cost-saving when varied in DSA. This indicates that the results are robust. Tirbanibulin is cost saving in 100% of the iterations in PSA and cost-effective in the majority of scenarios run through the model.

While the model was validated with Scottish clinicians at an advisory board, it is anticipated that the results of this economic evaluation are generalisable to the UK, whereby the patient pathway is anticipated to be similar to Scotland. While the efficacy data are generalisable to other countries, the resource use data and subsequent cost-effectiveness results are unlikely to be generalisable due to differences across treatment pathways worldwide. The predominant reason for this is the limited primary care involvement within AK treatment in some countries.

There have been no previously published economic evaluations of tirbanibulin for the treatment of AK, and few treatments for AK in general. One study analysed the cost-effectiveness of ingenol mebutate gel for AK from the perspective of the Scottish health system [29]; however, the license for this treatment has now been suspended [5]. Other analyses have considered photodynamic therapy and fluorouracil 5%, which appeared to be more costly than tirbanibulin in the current study [32, 38].

The modelling approach (structure and assumptions) used within the economic evaluation of ingenol mebutate gel is similar to the present study [29]. The present study includes additional comparators (imiquimod 3.75% and imiquimod 5%) and different parameters are used to inform the proportion of patients referred to secondary care (likely reflected by the change in clinical practice since the economic evaluation of ingenol mebutate gel was published in 2013). The authors are aware of three additional economic evaluations estimating the cost-effectiveness of AK treatments from a UK perspective [3, 32, 39]. All economic evaluations used complete clearance and/or response to inform the movement of patients through similar decision trees.

A key strength of this analysis is that an NMA was performed to determine the probability of treatment success associated with each comparator. NMA results are generally considered a robust data source as they allow differences in study design to be accounted for, minimising the potential for bias.

4.1 Limitations

The probabilities of AK recurrence and adverse events were obtained from targeted literature searches rather than NMAs due to a lack of RCT data. There is a risk of bias because no adjustment was made for potential differences in study design. However, recurrence was only modelled in a scenario analysis and adverse events were not a key driver of model results. The utility value associated with AK is highly uncertain due to the small number of patients within the study from which it was obtained (nine patients) [30]. Furthermore, it is not clear whether the utility data are generalisable to the Scottish population as they were elicited from a US population with a very small sample size. However, utility has a minimal impact on the cost-effectiveness results and are not key drivers of the model.

It was necessary to make simplifying assumptions when developing the model, largely due to a general paucity of data in this indication. The main assumptions were as follows: the probability of complete clearance was independent of the number of clinically visible lesions; complete clearance occurred at the time of maximum treatment response; no new lesions could develop for patients who experienced complete clearance; patients who did not experience complete clearance did not receive further treatment and lesions remained for the remainder of the model time horizon (i.e., lesion regression was not possible); and all disutilities associated with adverse events (severe local skin reactions) were applied for eight weeks. In addition, there was uncertainty regarding the proportion of patients referred to secondary care before treatment initiation and it was assumed that resource use would not differ depending on the number of lesions present due to the absence of published literature. All of these inputs and assumptions were validated by Scottish clinical experts and none changed the direction of model results when varied in DSA.

Furthermore, it was not possible to capture the progression of AK to SCC in the model due to a lack of relevant, up-to-date data regarding the probability of lesions becoming malignant. Progression to SCC has not been modelled in any previous economic evaluations of treatments for AK.

5 Conclusions

Tirbanibulin ointment is estimated to be cost saving and produces similar clinical outcomes at one year when compared to diclofenac sodium 3%, imiquimod 5% and fluorouracil 5% topical treatments. Despite the limitations and uncertainty associated with the model, in Scotland tirbanibulin is expected to be a cost-effective option for the treatment of AK on the face or scalp and remains cost-effective in the majority of sensitivity and scenario analyses presented. While the complete clearance rates are deemed similar across comparators, tirbanibulin is associated with a lower rate of severe local skin reactions and a shorter treatment duration, which may improve treatment adherence. These results enabled the SMC to advise that tirbanibulin is accepted for use within NHS Scotland.