Cost-effectiveness analysis trunk and limbs

National Clinical Guideline Centre (UK)

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

National Clinical Guideline Centre (UK). Psoriasis: Assessment and Management of Psoriasis. London: Royal College of Physicians (UK); 2012 Oct. (NICE Clinical Guidelines, No. 153.)

See 2019 Partial Update

Update information September 2017: The guideline has been revised throughout to link to MHRA advice and NICE technology appraisals that have been completed since original publication. Minor updates since publication August 2019: Links to the MHRA safety advice on the risk of using retinoids in pregnancy have been updated to the June 2019 version.

Update information September 2017: The guideline has been revised throughout to link to MHRA advice and NICE technology appraisals that have been completed since original publication. Minor updates since publication August 2019: Links to the MHRA safety advice on the risk of using retinoids in pregnancy have been updated to the June 2019 version.

Psoriasis: Assessment and Management of Psoriasis.

Show details

Contents

< Prev Next >

Appendix MCost-effectiveness analysis trunk and limbs

M.1. Introduction

The review of clinical evidence for topical therapies used in the treatment of individuals with mild to moderate plaque psoriasis showed that there were a wide variety of options – emollients, tars, dithranol, retinoids, corticosteroids (potent and very potent), vitamin D analogues and combination products – each associated with certain advantages and disadvantages. The results of the network meta-analysis indicated that some interventions, such as combined or concurrent vitamin D analogue and potent corticosteroid, were more likely to induce clearance or near clearance than others. Given that these combined and concurrent application strategies carry additional cost compared to both their individual constituent parts and compared to other topical alternatives, it is important to consider whether these additional costs are justified by additional health benefits in terms of improved quality of life.

Three cost-effectiveness analyses were identified in the published literature, but each had methodological limitations that called its conclusions into question. The analysis by Ashcroft and colleagues⁶⁰ was based on only one trial and included only two of the interventions of interest (dithranol and calcipotriol). The analysis by Oh and colleagues⁶¹ was quite old and had a fairly confusing model structure. The analysis by Bottomley and colleagues,⁶² although the most applicable of the included studies, used an unadjusted indirect comparison to inform the treatment effect estimates, which likely overestimated the effectiveness of some interventions and underestimated the effectiveness of others. Bottomley and colleagues also did not include all the possible comparators of interest.

Due to the limitations of the available economic evidence and the importance of this area in clinical practice, the GDG considered the development of an original cost-effectiveness model to evaluate topical therapies to be a high priority. The decision modelling presented here was developed in close collaboration between the health economist, NCGC technical team and GDG members.

M.2. Methods

M.2.1. Model overview

The analysis set out to evaluate the comparative cost-effectiveness of different topical therapy sequences used in the treatment of individuals with chronic plaque psoriasis. A cost-utility analysis was undertaken in line with the methods of the NICE reference case. QALYs were calculated using utility weights from EQ-5D responses and UK public valuations. Costs were considered from a UK National Health Service and Personal Social Services perspective and expressed in 2011 UK sterling. Healthcare costs associated with starting, maintaining and/or switching topical therapies as well as longer term costs of failing topical therapy were all included in the model.

The cost-effectiveness analysis must be relevant for decision-making over the longer term, as most people with psoriasis can be expected to require treatment for much of their lives. However, the evidence available for topical treatments is of short term duration and it would inappropriate to extrapolate for many years beyond treatment initiation given that the long term pathway of care is dependent on disease severity, access to specific facilities, patient preference and so on. Therefore, a 1-year time horizon was considered sufficiently long enough to capture the relevant costs and benefits associated with competing topical treatments.

To enable direct comparisons of treatments to be made based on the results of all relevant clinical trials, a network meta-analysis was performed and used to inform estimates of response (defined as clear or nearly clear) to treatment.

The performance of alternative treatment sequences was estimated using incremental cost-effectiveness ratios (ICERs), defined as the added cost of a given strategy divided by its added benefit compared with the next most expensive strategy. A threshold of £20,000 per QALY gained was used to assess cost-effectiveness.

All analyses were conducted probabilistically, thus capturing the imprecision and uncertainty around input parameter point estimates (i.e. mean/median odds ratios, utility weights, etc). A probability distribution was defined for various model inputs and when the model is run, a value for each input was randomly selected from its specific probability distribution simultaneously and costs and QALYs were calculated using these random values. The model is run repeatedly – in this case 5,000 times – and results are summarised as mean costs and mean QALYs. Probability distributions in the analysis were based on error estimates from data sources, such as confidence intervals. In addition, a series of one-way sensitivity analyses were run in order to test the effect of certain structural or variable uncertainties.

M.2.1.1. Comparators

The aim of the analysis was to identify the most cost-effective sequence of first, second and third line topical therapies. It was important to model sequences given that most patients will commence treatment with one topical and then try others before moving on to more intensive treatments such as phototherapy and/or systemic therapy. Table 8 presents the list of possible first, second and third line treatments which may be combined in a sequence.

Table 8

All possible sequences of first, second and third line interventions.

The following conditions were placed on the sequences, ensuring that they represented logical clinical practice:

Concurrent treatment with vitamin D analogue and potent corticosteroid would not come after a failure of once daily two-compound formulation product;
Once daily treatment with a given topical would not come after a failure of twice daily treatment with the same topical;
Once daily treatment with potent steroid or vitamin D analogue would not come after concurrent treatment with vitamin D analogue and potent corticosteroid or once daily two-compound formulation product;
No strategy could include potent corticosteroids among all three lines of treatment (including as part of concurrent vitamin D analogues and potent corticosteroid or TCF product).

Most comparators focus on evaluating a trial of three different treatments before referral for specialist review, but the GDG was also interested in whether earlier escalation of care might be more cost-effective. To test this, strategies have also been combined into two-treatment sequences with referral following a failure of second line treatment.

Due to the unacceptability of dithranol and coal tar as routine treatments (difficult application, risk of staining, strong and unpleasant odours, etc), these treatments were reserved for third line treatment only. This reflects their current placement in primary care given the availability of more acceptable and effective topicals such as those being compared as first and second line topicals.

M.2.1.2. Population

The analysis set out to evaluate the comparative cost-effectiveness of different topical therapy sequences used in the treatment of individuals with mild to moderate chronic plaque psoriasis.

M.2.1.3. Time horizon, perspective, discount rates used

The analysis took a UK National Health Service and Personal Social Services costing perspective, with costs expressed in 2011 UK sterling. A 1-year time horizon was considered clinically relevant and sufficiently long enough to capture important costs and consequences of first-line treatment in primary care. Since the time horizon was 1 year, no discounting rates were applied to either costs or benefits.

M.2.2. Approach to modelling

M.2.2.1. Model structure

A Markov model was constructed in TreeAge Pro 2009 to capture the different costs and effects associated with a given sequence of topical treatments. It was built to reflect transitions between a set of mutually exclusive health states, defined by response and non-response to treatment. The Markov model and how patients move through the pathway is illustrated in Figure 353. The structure of the model developed by the NCGC was adapted from the model developed by Bottomley and colleagues⁶² and was validated by the GDG as a reasonable reflection of current clinical practice.

Figure 353

Markov model of treatment with topical therapy.

The consequences of a given topical treatment are reflected as a set of possible transitions between health states over a series of discrete time periods, called cycles. In Figure 353, health states are depicted as ovals and interventions are depicted as rectangles. Movement between various health states is governed by transition probabilities, derived from the systematic review of clinical effectiveness data. Thirteen 4-week cycles were modelled, resulting in a 1-year time horizon for the analysis, with a half-cycle correction applied.

The model assumes that all hypothetical patients commence treatment with a given topical and experience one of two outcomes: response (defined as clearance/near clearance of their psoriasis) or no response (defined as something less than clearance/near clearance of their psoriasis). Patients who achieve clearance/near clearance are assumed to stop treatment and either maintain clearance/near clearance in the absence of treatment or they relapse. Patients who relapse are assumed to resume treatment with the same topical and again face a probability of responding or not responding. Patients who fail to achieve clearance on a given topical are assumed to return to their GP and receive a prescription for an alternative topical therapy.

Patients can receive up to three different topical therapies before being referred by the GP to a specialist review in an outpatient dermatology clinic where second-line treatment options could be considered. Some proportion of these referred patients will be kept on topical therapies, receive support and advice at the review consultation and be discharged back to their GP for long-term management. The remaining proportion will undergo a course of phototherapy and if they respond, they are discharged to their GP for long-term management.

M.2.2.2. Uncertainty

M.2.3. Model inputs

M.2.3.1. Summary table of model inputs

Model inputs were based on clinical evidence identified in the systematic review undertaken for the guideline, supplemented by additional data sources as required. Model inputs were validated with clinical members of the GDG. A summary of the model inputs used in the base-case (primary) analysis is provided in Table 9 below. More details about sources, calculations and rationale for selection can be found in the sections following this summary table.

Table 9

Summary of base-case model inputs.

Table 10

Overview of parameters and parameter distributions used in the model.

M.2.3.2. Baseline event rates

Creams and emollients with no active ingredient are a typical first-line therapy for patients presenting with plaque psoriasis. Although the primary objective of this model is to identify cost-effective sequences of topical therapies with active ingredients, it is useful to compare all strategies to a baseline probability of achieving clearance with a topical without an active ingredient. The absolute probability of achieving clearance or near clearance with twice daily vehicle/placebo was calculated by aggregating the number of people achieving clear/nearly clear across the twice daily vehicle/placebo arms of randomised controlled trials included in the systematic review of topical therapies and dividing by the aggregate sample size from the same arms. This resulted in a probability of 12.5% (95% CI: 10.4% to 14.6%) for achieving clear/nearly clear. For the probabilistic analysis, uncertainty in the risk parameter for vehicle/placebo was incorporated using a beta distribution (α=116; β=811).

M.2.3.3. Relative treatment effects

In order to estimate the effectiveness for all other comparators in the model, the treatment effect estimates from the network meta-analysis (see Appendix K) were applied to the baseline probabilities outlined above. In the base case, the estimates relating to the investigator assessed outcome (IAGI/PGA) were used. The effect estimates derived from the patient assessed outcome (PAGI) were used in a sensitivity analysis. In a further sensitivity analysis, the data from the network meta-analysis using all available data was used. The odds ratios used in the base case and each sensitivity analysis are presented in Table 11.

Table 11

Treatment effects.

To calculate the absolute probability of response to a given topical treatment, the odds ratio of that intervention compared to twice daily placebo from the network meta-analysis was converted into a relative risk and applied to the 12.5% baseline risk (e.g. probability of response to twice daily placebo) using the following formula:

P_T = P₀ × RR

Where P_T is probability or response to a given treatment; P₀ is baseline probability of response and

R R = \frac{O R}{1 - P_{0} (1 - O R)}

Where: OR is the odds ratio of the treatment compared to P₀, the baseline probability. The estimated probabilities of response for the base case and each sensitivity analysis are presented in Table 12.

Table 12

Probability of response.

For the probabilistic implementation of the analysis, uncertainty in the comparative treatment effects is incorporated by using 10,000 of the simulated odds ratios from the network meta-analysis. Using the simulated outputs allows us to preserve the joint posterior distribution from the network meta-analysis and any correlation of treatment effects.

Independent treatment effects were assumed across all interventions regardless of when they came in a sequence. In other words, the effectiveness of any topical as a second line intervention was not affected by what treatment may have come before.

Early versus late response

The data used to estimate the overall probabilities of response to treatment (Table 12) were based on trials of varying duration, 3 to 12 weeks follow-up. In the clinical review, we looked for evidence that would suggest when the appropriate time to assess response to treatment was. Where trials were of longer duration (i.e. 8 to 12 weeks) the evidence suggested that patients were still improving between 4 and 8 weeks. On that basis the GDG felt it would be inappropriate to assume that a) everyone who will respond will do so within 4 weeks and that b) patients who were not clear/nearly clear at the end of week 4 should discontinue treatment and be classified as a non-responders. Therefore, the model assumes that patients will be treated with a given topical for up to 8 weeks. If they respond in the first 4 weeks, then they are assumed to discontinue treatment. If they have not yet responded, then they are assumed to carry on for a further 4 weeks after which they discontinue having responded or not responded.

On that basis, where data from trials with longer follow-up was available, we looked to estimate what proportion of patients who responded by the end of follow-up had done so within the first 4 weeks or the last 4 weeks. The data with which to estimate this was patchy, but one trial²⁰ included our main 4 comparators (vehicle, potent corticosteroid, vitamin D analogue and two-compound formulation product) and reported response rates at both 4 weeks and 8 weeks. The data showed that a small proportion of people had responded to vehicle in the first 4 weeks, but by week 8 the number of responders was zero. On that basis, it was assumed that any response to placebo will occur in the first 4 weeks, with no additional responders in the following 8 weeks. For topicals with active ingredients, the data from Fleming 2010 indicated that of all responders to once daily vitamin D analogue at 8 weeks, one-third had achieved clearance by week 4. This figure was 57% and 59% for once daily potent corticosteroids and two-compound formulation product, respectively.

The proportions of early (0 to 4 weeks) and late (5 to 8 weeks) responders from Fleming 2010 were applied to the overall response figures generated from the network meta-analysis in order to estimate the probabilities of response in the first 4 weeks of treatment and the second 4 weeks of treatment (presented in Table 13). In the absence of data, the assumption was made that the proportions of early and late responders is the same for once and twice daily application of a given topical. In other words, this assumes that twice daily application of a topical does not induce response earlier than once daily application of the same topical. This assumption was validated by GDG member experience, which was that frequency of application did not have a demonstrable effect on speed of response.

Table 13

Probabilities of response: overall, early and late.

There was no trial data to inform the early compared to late responses for concurrent vitamin D and potent corticosteroid treatment, coal tar or dithranol. In the absence of data, the GDG made the assumption that the proportion of early and late responders to concurrent vitamin D and potent corticosteroid was likely to be the same as for potent steroid given that this is the component most likely to drive rate of response. For dithranol, graphs from Hutchinson 2000⁴¹ were judged to suggest that by the end of week 4, half of overall 8-week improvement in terms of IAGI and PASI had been achieved. Based on this, the assumption was made that the split between early and late response for dithranol was 50/50. Finally, in the absence of data, the GDG made the assumption that the early versus late breakdown for coal tar was the same as for dithranol.

M.2.3.4. Relapse

Psoriasis is a relapsing and remitting chronic condition and achievement of clearance/near clearance with active treatment has no long-term effect on the natural history of chronic plaque psoriasis. The RCT data with regard to relapse was quite sparse and inconsistent, due to a variety of factors including variable trial follow-up and differences in the definition of relapse. For the economic model, the GDG defined relapse as any deterioration to the point at which retreatment is required.

Given the lack of data, the GDG considered that there was little evidence to suggest any major differences between the proportions of patients relapsing or the time spent clear before relapsing following clearance with different topical treatments. The probability of relapse was set at 35.5% for all interventions and was varied in a sensitivity analysis. Average risk of relapse at 8 weeks follow-up across the trials where the outcome was reported was 58.4%. Uncertainty in this estimate for the probabilistic analysis was captured using a beta distribution (α=192; β=137). Assuming that the rate of relapse was constant over the 8 weeks, this translates to a 4-week risk of 35.5%.

It has been assumed that patients are at risk of relapse at any point following remission. In other words, patients who respond to treatment in the first 4 weeks of treatment may relapse within 4 weeks of discontinuing treatment or during any 4 week cycle thereafter.

Referral and specialist management

Sixty percent of hypothetical patients failing to respond to their third topical therapy are assumed to be referred for specialist review. This is based on figures quoted in the Dermatology Health Care Needs Assessment⁶⁴, which states that ‘although most patients have mild psoriasis, according to Nevitt and Hutchinson⁷¹, 60% had been referred for specialist care at some point.’ The 40 percent not referred onward are assumed to be managed by their GP for the time remaining in the model.

Among the 60 percent who are referred onward for consultation with a specialist, only 30% will be offered phototherapy. The other 70 percent will be given specialist advice and support about how to better manage their psoriasis with topical therapies. The 70/30 split used here is based on GDG opinion. In the GDG’s experience, the majority of patients who are referred to secondary care do not actually need more aggressive treatments like phototherapy or systemic therapy. They indicated that for around 70 percent of patients referred, topical therapy is likely to offer the best balance of efficacy and safety and that the goal of care at this point is to ensure patients know how and when to use topicals to maximise their efficacy. The model assumes that they receive this advice and support at one outpatient consultation and are then discharged back to their GP for long term management.

The 30 percent who receive phototherapy have a probability of responding based up on the results of the clinical evidence review presented in section 9.1 of the full guideline. The clinical evidence shows that around 86.7% of patients who receive a course of narrowband UVB (2 or 3 times weekly) will achieve clearance. For the probabilistic analysis, uncertainty in this estimate of effect was captured using a beta distribution (α=78; β=12).

M.2.3.5. Utilities

Achievement of clearance or near clearance and associated utility gain was used in the model to determine the impact of psoriasis treatment on overall health. Estimates of utility gain were taken from a recent cost-utility analysis included in the health economic review⁶². The mean utility at baseline was 0.8 and mean utility gain associated with clearance/near clearance was 0.09. It is expected that patients who do not achieve clearance or near clearance will still experience some level of improvement on treatment; therefore, these patients also experience a modest utility gain. Bottomley and colleagues modelled a utility gain of 0.07 for non-responders, but the GDG considered this to be optimistic. They felt that the difference between responders and non-responders was likely to be greater, and therefore recommended a utility gain for non-responders compared to baseline to be slightly less, at 0.05. Due to the uncertainty in this parameter, it was varied in sensitivity analysis.

Table 14

Health state utility values.

Key assumptions about utilities in the model:

Patients who do not achieve clearance at 4 weeks and continue on for a further 4 weeks of topical therapy will improve somewhat and therefore accrue the gain associated with non-responders.
Patients who relapse following clearance lose the incremental gain between response and non-response (0.04) before resuming treatment.
Patients who fail to respond and ultimately reach the point of requiring referral to a specialist or phototherapy return to their baseline level of utility (0.8).
Patients managed long-term by either a GP or a specialist accrue the gain associated with non-responders.

M.2.3.6. Resource use and cost

Topical therapy

Resource use of alternative topical treatments was based on reported mean quantities of study drugs used by patients in the RCTs²¹^,²⁷^,³⁸^,⁶⁸ at the end of 4-week treatment periods. No estimates were available to inform the mean usage of coal tar used twice daily. In the absence of data, we assumed that the mean usage for coal tar would be approximately equal to that of dithranol. No estimate from an RCT was available to inform the mean quantities of vitamin D analogue and potent corticosteroid when they are used concurrently (e.g. one in the morning and the other in the evening). In the cost-utility analysis by Bottomley and colleagues⁶², they estimated mean usage for this strategy to be 160.9 g (95% CI: 140.7–181.1) based on an unpublished trial they held on file. We have taken this estimate for use in our model, assuming that the total usage is split evenly between vitamin D analogue and potent corticosteroid. Mean quantities and distribution parameters for the probabilistic analysis are presented in Table 15.

Table 15

Mean quantities of topicals used per 4-week cycle.

Unit costs of topicals (Table 16) were taken from the most recent BNF⁷². Given that the interventions were modelled assuming a class effect, the cost of topical had to be selected from a variety of compounds, formulations and package sizes. For simplicity, we used the cost for the topical with the lowest unit cost per gram/millilitre.

Table 16

Unit costs of topical therapies.

To calculate the per cycle cost of each topical, the mean quantities were converted into the cheapest combination of the number of packs of topical needed. For example, the mean 4-week dosage for twice daily potent corticosteroids was 144.5 g. The cheapest combination of packs needed to provide this quantity was one 100 g pack and two 30 g packs. The 4-week costs of topical treatment based on the mean quantities used are presented in Table 17.

Table 17

Mean cost of 4-week topical treatment.

During probabilistic implementation, dosages were drawn from topical specific gamma distributions fitted using the mean reported in the RCTs and a standard error assumed to be 20% of the mean. The model was built to ensure that the cheapest combination of packs, as outlined in the example above, could be calculated automatically for any sampled value. For example, if the sample value for twice daily potent corticosteroid was 180 g, then the cheapest combination would be automatically be calculated as two 100 g packs. Similarly, if the sampled value was 45 g, then the cheapest combination would be two 30 g packs.

Dithranol was assumed to be titrated up over the course of the first 4-week cycle, starting with 0.1% strength for the first week, followed by 0.25%, then 0.5% and finally 1%. The total dosage over the 4-week period was assumed to be distributed equally between the different strengths.

A different costing method was used for twice daily vehicle. Because the vehicle cream comes in large packs (500 g), the cost was applied per gram used during a 4-week cycle instead of per pack used during a 4-week cycle.

Health care consultations

It was assumed that following a failure (non-response) of a given topical treatment, patients returned to their GP for review and receive a second or third topical or referral for specialist review. Thus, each change in topical treatment will accrue a cost of a GP visit. Patients experiencing a relapse following successful treatment with a given topical are assumed to get a repeat prescription for the same topical without accruing the cost of a GP visit.

Sixty percent of patients who fail to respond to a third topical treatment are referred by their GP for specialist review. During the time spent between being referred and the specialist review, patients are assumed to maintain topical treatment, for which the average 4-week cost across all topical treatments was used (£29.78).

Each patient who is referred is seen by a consultant dermatologist in an outpatient clinic, thus accruing this cost. Based on GDG experience, it was assumed that 70% of these referred patients will be kept on topical therapies, receive support and advice at the review consultation and be discharged back to their GP for long-term management. The other 30% are assumed to undergo a course of phototherapy, thus accruing the cost of 24 sessions of narrowband UVB. Responders to narrowband UVB are assumed to be discharged to their GP for long-term management; non-responders are assumed to be managed in long-term specialist care.

In reality, some of this 30% referred for phototherapy might attend a day centre where they would undergo treatment with specialist applied topicals such as dithranol and crude coal tar. For reasons of pragmatism and simplicity, this alternative on the clinical care pathway was excluded from the base case. However, in a sensitivity analysis, we added in the likely costs of such treatments in order to observe how the results might change.

Table 18

Unit cost of health care consultations.

M.2.4. Computations

The model was constructed in TreeAge Pro 2009 and was evaluated by cohort simulation. All hypothetical patients start treatment with a topical therapy and either achieve clearance or near clearance or do not. Following the achievement of clearance/near clearance, patients can subsequently relapse and upon resumption of the same topical therapy either respond or do not respond and move on to the next topical therapy in the sequence. Movement between health states in subsequent cycles is determined by the various probabilities described in the preceding sections. Each 4-week cycle the cohort spends in a given health state is counted.

Total QALYs were calculated from the above information as follows. Each 4-week cycle, the time spent in each health state of the model was weighted by the utility for that state. The QALYs per cycle were then discounted to reflect time preference. QALYs during year one were not discounted. The total discounted QALYs was the sum of the discounted QALYs per cycle.

T o t a l d i s c o u n t e d Q A L Y s = \sum_{t = 1}^{i} \frac{Q (t)}{{(1 + r)}^{t - 1}}

Where: t=cycle number; i=maximum cycle number; Q(t) = QALYs in cycle t; r = discount rate

Total costs were calculated from the above information as follows. Each cycle, the time spent in each state of the model was multiplied by the costs for that state. The costs per cycle were then discounted to reflect time preference. Costs during year one were not discounted. The total discounted costs were the sum of the discounted costs per cycle.

T o t a l d i s c o u n t e d c o s t s = \sum_{t = 1}^{i} \frac{C (t)}{{(1 + r)}^{t - 1}}

Where: t=cycle number; i=maximum cycle number; C(t) = costs in cycle t; r = discount rate

The used cost-effectiveness metric is the incremental cost-effectiveness ratio (ICER). This is calculated by dividing the difference in costs associated with two alternatives by the difference in QALYs. The decision rule then applied is that if the ICER falls below a given cost per QALY threshold, the result is considered to be cost effective. If both costs are lower and QALYs are higher, the option is said to dominate and an ICER is not calculated.

I C E R = \frac{C o s t s (B) - C o s t s (A)}{Q A L Y s (B) - Q A L Y s (A)}

When there are more than two comparators, as in this analysis, options were ranked in order of increasing cost and then options ruled out by dominance (i.e. those that were more costly and less effective than alternate strategies) or extended dominance (i.e. where a linear combination of other strategies could produce greater benefit at lower cost) were excluded before calculating ICERs. ICERs were calculated based on mean costs and effects as estimated during the probabilistic implementation of the model.

The effect of uncertainty in the results is reflected by the reporting of 95% confidence intervals around mean total costs and effects. Secondly, uncertainty was illustrated by estimating the probability a given AED was the optimal treatment option. For strategy X, this was calculated as

Net Benefit(X) = (QALYs(X) × D) − Costs(X)

Where: Costs/QALYs(X) = total discounted costs/QALYs for option X; D=threshold

The decision rule then applied is that the strategy with the greatest net benefit is the cost-effective option at that threshold. That strategy is expected to provide the highest number of QALYs at an acceptable cost. The probability a given AED is optimal is calculated as the proportion of simulations where that option had the greatest net benefit at the specified threshold.

M.2.5. Sensitivity analyses

A series of one-way sensitivity analyses and scenario analyses were performed to assess how changes in one or more parameters or assumptions might change the conclusions of the analysis. In one set of sensitivity analyses, alternative estimates of treatment effects from the network meta-analyses (Appendix K) were used. In a second sensitivity analysis, the utility value associated with non-response was varied upward to match the estimate used by Bottomley and colleagues⁶². In a third set of sensitivity analyses, the quantity of TCF product used over a 4 week treatment period was reduced to match the estimate used by Bottomley and colleagues. In a fourth series of sensitivity analyses, estimates of future resource use and cost were altered and the time horizon was lengthened. Finally, alternative assumptions about the comparators were used to explore what might be appropriate if there were concerns about safety or contraindications.

M.2.6. Model validation

The model was developed in consultation with the GDG; model structure, inputs and results were presented to and discussed with the GDG for clinical validation and interpretation.

The model was systematically checked by the health economist undertaking the analysis; this included inputting null and extreme values and checking that results were plausible given inputs. The model was peer reviewed by a second experienced health economist from the NCGC; this included systematic checking of the model calculations.

M.3. Results

M.3.1. Base case

This analysis found that, given a NICE willingness-to-pay threshold of £20,000 per QALY gained, the most cost-effective strategy is likely to be one of starting with twice daily potent corticosteroid and moving to concurrent potent corticosteroid and vitamin D analogue and then twice daily coal tar. This strategy was also the least costly strategy among the 118 modelled. Base case results for non-dominated and non-extendedly dominated strategies are presented in Table 19.

Table 19

Incremental analysis of base case results – psoriasis of trunk and limbs.

Results showed that starting with concurrent potent corticosteroid and vitamin D analogue and switching to twice daily potent corticosteroid and then twice daily coal tar is £9 more costly over 1 year and only produces 0.00041 more QALYs than the least costly strategy mentioned above. This gives it an incremental cost-effectiveness ratio (ICER) of £22,658 which is just above the NICE £20,000 per QALY threshold.

The most effective strategy (once daily TCF then twice daily potent corticosteroid then twice daily coal tar) costs an additional £192 per year compared to the next most costly non-dominated strategy (concurrent steroid and vitamin D then twice daily potent steroid then twice daily coal tar), yet produces just 0.00107 additional QALYs for an ICER of over £179,000. Based on the results of this model, it appears that starting with once daily TCF, although most effective, is very unlikely to be cost-effective.

Mean costs and QALYs and their respective 95% confidence intervals for all strategies, ranked in order of mean net benefits at a £20,000 per QALY threshold, are presented in Table 20. These show that a strategy of using vehicle or emollient with no active agent only was the most costly and least effective, largely driven by the cost of referrals and specialist management for non-responders. Strategies that included once or twice daily vitamin D were not cost-effective regardless of where they were included in the sequence. This is largely due to their relatively low rank in terms of effectiveness and their relatively high acquisition cost. Strategies that included dithranol were also all dominated, that is more costly and less effective than alternatives. Finally, strategies in which patients were referred after non-response to only 2 topicals were all dominated, thus not cost effective.

Table 20

Mean total costs and QALYs for all modelled comparators.

A breakdown of total costs by type of resource use (i.e. topicals, GP visits, outpatient consultations, phototherapy) is presented for all modelled strategies in Table 21. Note that these costs were produced by a deterministic run of the model and therefore may not match exactly the total costs presented from the probabilistic analysis in Table 20; however, they are very similar. Disaggregation of costs allows one to observe what part of a given strategy is driving the majority of total cost. Strategies that are less effective tend to have higher downstream costs driven by visits to the GP and referrals for specialist review and/or phototherapy. Strategies that are very effective are likely to have lower downstream costs, but potentially higher drug costs. Based on this disaggregation, it becomes clear that strategies with TCF product or vitamin D analogue have relatively high topical costs, some of which are offset by reduced downstream costs in terms of consultations with specialists and courses of phototherapy. Strategies with potent corticosteroids offered alone or in combination with vitamin D analogue (concurrent therapy) show similar downstream costs as strategies involving TCF product, but because their acquisition cost is dramatically lower, the overall total cost is much lower.

Table 21

Disaggregated total costs by items of resource use.

The probabilistic analysis indicates that there is a great deal of uncertainty as to which sequence is optimal (i.e. most cost effective). There appears to be very little difference between initial potent corticosteroid followed by concurrent corticosteroid and vitamin D and vice versa, with the difference in their net monetary benefits (NMB) being only £1 (£16,748 and £16,747 respectively) and both having an equal probability of being optimal at a £20,000 willingness to pay threshold. Generally, it looks as though a strategy of starting with either potent corticosteroids or concurrent treatment with potent corticosteroid and vitamin D analogue is most likely to be cost-effective, whereas starting with once daily TCF product is very unlikely to be cost-effective.

M.3.2. Sensitivity analyses

A series of sensitivity analyses suggested that the conclusions from the base case are somewhat sensitive to changes in some parameters and/or assumptions.

M.3.2.1. Treatment effects

The network meta-analysis of topical therapies was performed for two response outcomes: investigator assessed global improvement (IAGI) and patient assessed global improvement (PAGI). The economic evaluation used the investigator assessed outcome in the base case, largely because there was more data from the randomised evidence reported for this outcome. In a sensitivity analysis, treatment effects from the network meta-analysis of patient reported outcome was used. Results of this sensitivity analysis are presented in Table 22.

Table 22

Incremental analysis of sensitivity analysis using patient-reported outcome (PAGI).

Results of the analysis using patient reported outcomes indicates that starting treatment with once daily potent corticosteroids, moving on the concurrent treatment if that fails and then trying twice daily vitamin D analogue is likely to be both the least costly and most cost-effective strategy given a threshold of £20,000 per QALY gained. Initial treatment with concurrent potent corticosteroid and vitamin D analogue appears less cost-effective using patient reported outcomes than physician reported outcomes, unlikely to be cost-effective at thresholds less than £100,000. Once daily TCF product, first or second line in a sequence, still looks to generate additional benefits (QALYs), but at additional costs unlikely to be considered good value for NHS resource (ICERs upwards of £115,000 per QALY gained).

The base case network meta-analysis of physician/investigator assessed response used in the base case cost-effectiveness analysis included all RCTs that met the inclusion criteria for the clinical review of direct evidence. The review of direct evidence was quite focused and as such did not include evidence for every possible pair wise comparison. In a sensitivity analysis of the network meta-analysis and thus the cost-effectiveness analysis, additional studies were included. For details on the particulars of these sensitivity analyses and what effect they had on the estimated treatment effects, see Appendix K.

When treatment effects were based on all relevant RCT data, the results of the base case changed only slightly. Twice daily potent corticosteroid followed by concurrent steroid and vitamin D analogue is still likely to be optimal for first and second line treatments. However, instead of twice daily coal representing the optimal third line topical, twice daily vitamin D analogue looks to be most cost-effective. This sensitivity analysis calls into question whether vitamin D or coal tar represents the better third line treatment option.

M.3.2.2. Variation in early versus late response

The base case assumed that patients would trial a given topical for up to 8 weeks and that some proportion of patients would be expected to respond by 4 weeks and discontinue treatment at that time. The remainder would carry on to 8 weeks, at which time non-responders would move on to the next topical in a sequence. The data defining the breakdown of early (at 4 weeks) vs late (at 8 weeks) responders was limited to two studies⁴¹^,⁷³ and GDG opinion and was thus very uncertain. Deterministic sensitivity analyses were performed around these parameters to observe the impact on the results.

First, an analysis was performed in which no one was expected to respond and discontinue treatment at 4 weeks (i.e. all responders require 8 weeks treatment). Compared to the results of the base case when all comparators are included, the rank order of strategies in terms of mean net benefits changed very little. The ICERs for strategies on the cost-effectiveness frontier (see Table 19) increased relative to the base case, thus becoming less likely to be considered cost-effective.

Second, an analysis was performed in which all responders were assumed to respond by 4 weeks, with no one requiring an additional 4 weeks of treatment. The ICER for all strategies on the cost-effectiveness plane (see Table 19) decreased relative to the base case, and now starting with concurrent therapy and moving to twice daily potent corticosteroids looks to be cost-effective at a £20,000 threshold compared to potent corticosteroids and then concurrent therapy. Initial treatment with once daily TCF product is still unlikely to be cost-effective, with an ICER of more than £140,000.

Finally, an analysis was performed in which a 4-week stopping rule was applied. In this scenario, responders were limited to those that have responded by week 4 (see Table 13), and all other patients are assumed to move on to the next topical in the sequence (i.e. no one continues to 8 weeks of treatment with the same topical). Relative to the base case, the total costs for all strategies more than doubled as more patients were classified as non-responders and moved down the care pathway reaching referral to secondary care. Starting with concurrent therapy and then moving to twice daily potent corticosteroids was now the least costly strategy and most likely to be cost-effective. The ICER for once daily TCF product instead of concurrent therapy in this sequence decreased substantially relative to the base case (£174,000 to £94,000) but is still unlikely to be considered cost-effective at the NICE threshold.

M.3.2.3. Reduced adherence

There was some concern that issues of treatment adherence were inadequately captured in the model. The estimates of effect used in the base case were derived from randomised controlled trials which may represent the best case scenario for topical therapies. The GDG wished to explore how reduced adherence to twice daily treatments would affect the conclusions of the base case. In this scenario, 60% of patients being treated with twice daily topical were assumed to adhere to twice daily treatment whilst the remaining 40% of patients were assumed to apply the topical only once daily⁷⁴. For concurrent therapy, the 40% were assumed to adhere to once daily potent corticosteroid treatment only. Efficacy of the twice daily treatments would thus be reduced compared to the base case estimates. To be conservative, no reductions in cost were assumed despite the fact that less topical would be used.

With adherence reduced, there is no change substantive change to the results of the base case. Total costs across all strategies increase slightly (average of £27 more) and benefits decreased very slightly (average of 0.0007 fewer QALYs), but the conclusions from the base case remain unchanged. The most cost-effective strategy, given a £20,000 per additional QALY threshold is still twice daily potent corticosteroid followed by concurrent therapy and then twice daily coal tar. To put concurrent therapy before twice daily potent corticosteroids has an ICER of £36,000 (up from £23,000 in base case) and to replace concurrent therapy with once daily TCF before steroids has an ICER of £76,609 (down from £174,545 in the base case).

M.3.2.4. Utility values

In the base case, the mean utility gain associated with achieving some level of improvement, but not clearance or near clearance was assumed to be 0.05. This value was based on a downward adjustment of a value used in a recent cost-utility analysis included in the health economic review. Bottomley and colleagues⁶² modelled a utility gain of 0.07 for non-responders compared to baseline. To see what effect the GDG adjustment had on the results, the Bottomley figure (0.07) was used in a sensitivity analysis

Results indicate that the conclusion about cost-effectiveness changes very little using this more optimistic estimate of utility gain. The ICERs for all strategies increases relative to the base case; therefore, starting with concurrent treatment before twice daily potent corticosteroids is less likely to be cost-effective (ICER=£88,333 vs £23,250 in the base case). Similarly, the ICER for a strategy starting with TCF product increased to over £787,000 compared to starting with concurrent treatment (£174,500 in the base case).

M.3.2.5. 4-week quantity of TCF product

In the base case, hypothetical patients are assumed to use 134.0 g of TCF product during 4 weeks of treatment. Bottomley and colleagues used a much lower value for this input (92.6 g), and we explored how the results of the NCGC analysis might change if this lower estimate was used. The cost of 92.6 g of TCF product was £61.27 (compared to £94.26 in the base case). The results of this sensitivity analysis showed that the ICER for TCF product improved compared to the base case (£124,400 vs £174,545); however this is still well above the NICE cost-effectiveness threshold of £20,000 per additional QALY. Initial therapy with twice daily potent corticosteroid or concurrent vitamin D analogue and potent corticosteroid is still more likely to be considered cost-effective.

M.3.2.6. Unit costs of potent corticosteroids and vitamin D analogues

The base case assumed that the cost for each topical was based on the product and formulation with the lowest unit cost per gram/millilitre. Given that clinicians and patients may have preferences for different products or formulations, it was considered necessary to explore how variation in price of topicals, particularly potent corticosteroids and vitamin D, might affect the results. To do this, the highest cost (per gram) potent corticosteroid Synalar gel (fluocinolone acetonide) was assumed in place of Betnovate cream or ointment. The cost of Synalar gel is around four times that of Betnovate cream/ointment. In another analysis, the most costly vitamin D ointment, Curatoderm (tacalcitol), was assumed instead of Silkis (calcitriol). The cost of Curatoderm is around 2.5 times more costly than Silkis and 1.6 times more costly than Dovonex (calcipotriol) ointment. In a final sensitivity analysis, both Synalar gel and Curatoderm were used. Results in terms of incremental cost-effectiveness ratios are presented in Table 23.

Table 23

Incremental cost per QALY gained under different treatment cost assumptions.

When the cost of Synalar gel is used, the ICER for starting with concurrent therapy and then moving to potent corticosteroid compared to the reverse, decreases substantially from the base case (£4,365 compared to £23,250), becoming optimal given the NICE threshold. The ICER for this strategy when only the cost of Curatoderm ointment is used and when Synalar gel and Curatoderm ointment, actually increase relative to the base case. Even with increased costs for potent corticosteroid and vitamin D, once daily TCF product is unlikely to be cost-effective compared to concurrent therapy unless the willingness to pay threshold is well over £100,000 per QALY gained.

M.3.2.7. Sensitivity analyses – Restricted comparators

The base case analysis put several conditions on the way topicals could be sequenced (see M.2.1.1). These conditions did not restrict how potent corticosteroids were fit into treatment sequences other than that they could not appear in all three lines of treatment. This included their use as part of concurrent or combined (TCF product) treatment. The GDG expressed concern that these restrictions may not fully reflect the caution they would use in prescribing trials of potent corticosteroids, in that the BNF discourages continuous use of potent corticosteroids for more than 8 weeks at a time. The GDG was also concerned that the analysis did not fully capture the safety risks associated with the continuous or intermittent use of twice daily potent steroids. In a series of sensitivity analyses, various additional restrictions were placed on the treatment sequences.

In the first scenario, it was assumed that interventions that included potent corticosteroids could not be offered consecutively. For example, once daily TCF product could not be offered after treatment with once or twice daily potent corticosteroids, nor could twice daily potent corticosteroid follow once daily potent corticosteroid. Under this assumption, starting with twice daily corticosteroid, then trying twice daily vitamin D analogue and then using both concurrently would represent the best value for NHS resources given a £20,000 per QALY threshold. Starting with concurrent treatment would only be cost-effective at thresholds of greater than £33,000 and TCF product would only be cost-effective at thresholds over £202,000.

In the second scenario, it was assumed that twice daily corticosteroid could not be prescribed as a first or second line topical therapy, but consecutive use of potent corticosteroids was permitted. Under this scenario, the optimal strategy was to start with concurrent corticosteroid and vitamin D analogue, then try twice daily vitamin D analogue alone and finally twice daily potent corticosteroid only. This had an ICER of £18,000 per QALY gained compared to once daily potent corticosteroid followed by concurrent treatment and then twice daily coal tar. Strategies including TCF product either as second or first line were not cost-effective unless the threshold was over £110,000 and £446,000, respectively.

A third scenario combined the first and second scenarios, such that twice daily potent corticosteroid could not be prescribed as first or second line treatment and no sequences could include consecutive lines of potent steroid containing strategies. Under these conditions, the same sequence as in scenario 2 is most cost-effective (Concurrent – vit D BD – PS BD). TCF product replaces twice daily steroid in that sequence only if the threshold willingness to pay is £134,000 and replaces concurrent treatment in the same sequence if the threshold is £202,000.

In a fourth and final scenario, twice daily potent corticosteroid was removed entirely and no potent steroid containing products could be prescribed consecutively. Under this assumption, the most cost-effective sequence was initial concurrent treatment followed by twice daily vitamin D alone and then twice daily coal tar. TCF product replaces twice daily coal tar in that sequence at a threshold of over £47,000 and replaces concurrent treatment at a threshold of over £489,000.

Results from all aforementioned sensitivity analyses (i.e. treatment effects, early versus late response, reduced adherence, cost of potent corticosteroids and vitamin D and so on) were reinterpreted within the context of these restricted comparator scenarios. The conclusions from each scenario presented here were insensitive to changes in the tested parameters. For example, concurrent therapy followed by twice daily vitamin D followed by twice daily potent corticosteroids was optimal across all tested parameter variation under the conditions that twice daily potent corticosteroids could not be offered as initial treatment or when steroids could not be used consecutively. Furthermore, once daily TCF product was consistently more effective but never found to have an ICER below or near to the NICE £20,000 per QALY threshold.

M.3.2.8. Downstream resource use and cost

Changes to the assumed probability of referral to secondary care and proportion offered phototherapy have no meaningful effect on the conclusions of the base case. The probability of referral to secondary care was varied downwards to 40% and upward to 80%. When referral occurred less often than in the base case, there was no change to the rank order of strategies, but the ICER for a strategy where TCF product was used first instead of concurrent treatment increased to £200,000 per additional QALY. When referral occurred more often than in the base case, there was still no change in the rank order, but the ICER for TCF product was slightly lower. If the probability of undergoing UVB phototherapy upon referral was higher than in the base case (50% vs 30%), then the ICER for TCF product compared to concurrent treatment reduced slightly, but not enough to make it cost-effective. Finally, if instead of assuming patients are treated with UVB phototherapy, it is assumed they receive outpatient day care treatment with specialist supervised topical therapies, then the ICER for concurrent therapy before potent corticosteroids alone increases to over £30,000 per QALY and the ICER for initial TCF product instead of concurrent therapy decreases to £155,000 per QALY.

If the time horizon is extended for 2 to 3 years and cumulatively more patients see a specialist and move on to UVB phototherapy, then initial treatment with concurrent vitamin D and potent corticosteroids becomes more cost-effective than starting with potent corticosteroids alone. When the time horizon is extended, TCF product becomes more cost-effective compared to concurrent treatment (ICER = £118,067 at 2 years; ICER = £90,710 at 3 years; ICER=£75,255 at 5 years; ICER=£73,541 at 10 years), but is still very unlikely to be considered cost effective given the NICE willingness to pay threshold of £20,000 per QALY gained. Visual inspection of the health state membership probabilities over a 10-year time horizon indicates that patients are no longer transitioning between health states after 8 years because they have all reached long-term management with a GP or specialist by this point. This suggests that the ICER for TCF product is unlikely to come down any further even if the model time horizon is extended beyond 10 years.

M.4. Discussion

M.4.1. Summary of results

In assessing the relative cost-effectiveness of alternative topical therapies in patients with mild to moderate psoriasis limited evidence was available from the published economic literature. The evidence that was identified and included in the health economic review had potentially serious limitations and therefore the GDG considered it a priority to undertake original evaluation for the guideline in order to inform recommendations. This analysis showed that there were relatively small differences in terms of benefit between different topical sequences, but the differences in terms of cost were quite substantial. Based on the mean costs and benefits of 118 compared sequences, the analysis suggests that initial treatment with potent corticosteroids followed by concurrent treatment with potent corticosteroid and vitamin D analogue (morning/evening application) and followed then by twice daily coal tar therapy is likely to represent the most cost-effective sequence for implementation in primary care. Uncertainties in the analysis were explored through sensitivity analysis which showed that in some scenarios

Once daily potent corticosteroid or concurrent treatment should come first in the sequence
Twice daily vitamin D analogue should come second or third in the sequence, after concurrent treatment
TCF product should be offered third in the sequence, after potent corticosteroids and concurrent treatment

Sequences starting with once daily TCF product were slightly more effective than the same sequence starting with concurrent potent corticosteroid and vitamin D analogue; however, the very modest additional benefit (0.0011) would only be considered potentially cost-effective if willingness to pay thresholds were between £100,000 and £500,000 per QALY gained.

M.4.2. Limitations & interpretation

The analysis has several limitations which were considered carefully by the GDG. Firstly, the analysis evaluates treatment sequences even though the available trial data compares single topicals head to head without sequencing. In order to apply the treatment effects within the sequencing model, we assumed that treatment effects were independent. That is, we assumed the effectiveness of TCF product as a second or third line topical was equal to its effectiveness as a first line agent and that this was true regardless of other topicals it may follow. The GDG did not believe this to be a significant limitation given that the patients included in the overwhelming majority of RCTs were reported to have psoriasis for longer than 5 years, during which the can be assumed to have previously tried, succeeded and/or failed various topical treatments.

The analysis only captured the efficacy of topicals and did not capture the costs or consequences of adverse events. Although the RCT evidence on adverse events was sparse, the GDG is aware of the risks associated with the long-term use of potent and very potent corticosteroids. They carefully considered whether the added effect in terms of clearance was worth the potential risks of adverse effects.

The model was also focused on the induction of disease clearance as opposed to the maintenance of clearance. Trials focusing on maintenance were limited in number and inadequately reported for use in the economic model. In particular, there was uncertainty as to how maintenance treatments were applied in the trials and therefore incorporating such evidence and assumptions into the model was considered too difficult and unlikely to be valid.

The model also takes a relatively short time horizon considering that psoriasis is a chronic, long term condition for which patients may undergo treatment for many years of their lives. Frequency and severity of relapse, selection for and speed of onward referral, methods of self-management and long-term safety are all issues inadequately addressed in the evidence base and therefore translate into limitations of the economic analysis.

The model estimated the health gain for each treatment by mapping the change in PASI score to the EQ-5D based on observational evidence. However, it has been noted that several important areas of health-related quality of life for people with psoriasis are not directly assessed by the EQ-5D questionnaire⁷⁵. Therefore it is possible that the EQ-5D may lack content validity for these patients. Research is ongoing in this area. But we note that even using a £30,000 per QALY threshold rather than £20,000 would not change the conclusions of our analyses. Therefore only if the EQ-5D is under-estimating health gain of one treatment compared to another by a considerable extent, could this pose a serious limitation.

The analysis specifically found twice daily potent corticosteroid to be highly cost-effective, but the GDG expressed concern that the well known side effects of potent corticosteroids (e.g. skin atrophy, rapid relapse) were not adequately captured in the economic model owing to a lack of data. Twice daily potent corticosteroids came out more cost-effective than once daily, largely because the quantities of topical used for once and twice daily application were very similar, yet the network meta-analysis showed a non-significant trend toward twice daily being more effective in the investigator assessed outcomes used in the base case (OR=1.807, 95% CrI 0.42 to 8.07). However, this trend is reversed for the patient assessed outcome – twice daily performed less well than once daily (OR=0.714, 95% CrI 0.14 to 3.55). This finding is reflected in the results of this sensitivity analysis where patient reported response was used, which show once daily to be more cost-effective than twice daily. The consensus of the GDG was that they could not be certain that twice daily potent corticosteroids were more effective than once daily potent corticosteroids. They concluded that even if twice daily application was more effective at inducing clearance or near clearance than once daily application, the risks of higher dose steroids were very likely to outweigh the potential benefits and make the intervention comparatively less effective and cost-effective. Therefore the GDG excluded strategies that included twice daily corticosteroids in the first two lines of treatment. It was considered appropriate as third-line treatment, as the number of patients exposed to the risks would be fewer but the need for efficacy more urgent. In order to avoid continuous treatment with steroids for more than 8 weeks the GDG also chose to exclude strategies that contained corticosteroids in two consecutive lines of treatment. After these considerations the most cost-effective strategy was:

1^st line – Concurrent treatment with potent corticosteroid and vitamin D analogue (morning/evening application)
2^nd line – twice daily vitamin D analogue
3^rd line – twice daily potent corticosteroid

The GDG specifically considered whether they should offer concurrent treatment (morning/evening) with two separate topicals or offer combined treatment in a single product for use just once daily. They considered the results of the cost-effectiveness analysis which showed that combined treatment (once daily TCF product) is not cost-effective compared with concurrent treatment. This is because the network meta-analysis found them to have similar efficacy, but TCF product is much more costly (unit cost of 120 g combined product containing calcipotriol monohydrate and betamethasone dipropionate is between 2 and 4 times more costly than combined unit cost of 100 g of vitamin D analogue and potent corticosteroid each). This is true even when the most costly potent corticosteroid and vitamin D products and formulations are assumed to be prescribed. The GDG considered whether a once daily application of the combined product may be cost-effective when considering the problems many patients have adhering to twice daily treatment regimens. The results of a sensitivity analysis wherein 40% of patients prescribed concurrent therapy were assumed to apply only their potent corticosteroid once per day showed that the very small benefits of once daily combined product were still outweighed by its extra cost. The GDG concluded that the combined formulation product as first-line treatment produced enough additional benefit to justify its substantial additional cost.

The base case cost-effectiveness analysis and sensitivity analyses showed that the choice of third line treatment in a given sequence was highly uncertain. Depending upon the data used and assumptions made, third line treatment with twice daily potent corticosteroid, twice daily coal tar, or once daily TCF product was likely to be most cost effective. To reflect the uncertainties in the conclusions about cost-effectiveness and provide prescribers and patients with a degree of choice, the GDG chose to recommend all of these interventions if the patient has failed to achieve clearance or near clearance with Concurrent treatment with potent corticosteroid and vitamin D analogue (morning/evening application followed by a course of twice daily vitamin D analogue. They considered that some people may not choose to use coal tar as it has a pungent odour and that some people may prefer vitamin D analogues as they are generally safe for long term use. They considered that the combined potent corticosteroid and vitamin D analogue product was much more costly than other alternatives, but it may represent value for NHS resource in a select group of patients with resistant mild to moderate psoriasis. It also may be more cost-effective to offer if the alternative is referral and escalation of treatment to much costlier interventions (e.g. phototherapy, specialist applied topicals, systemic therapy, biologic therapy).

The NCGC cost-effectiveness did not find short contact dithranol to be more cost-effective than other first, second and third line alternatives in the base case or any sensitivity analyses. The GDG did not want to rule dithranol out as a treatment option for some patients, but considered it only potentially cost-effective for patients who have failed to respond to other more efficacious and easy-to-use topical therapies. They emphasised the need for health care professional to clearly explain proper application of dithranol for home use in order to maximise its effectiveness and reduce the inconvenience. They also considered that dithranol may be best delivered as part of treatment in a day care setting with specialist nurse supervision.

The cost-effectiveness of very potent corticosteroids was not evaluated as part of the NCGC decision modelling as the GDG did not consider it to represent a safe treatment option for the management of mild to moderate psoriasis being managed in primary care. They considered that based on its efficacy and relatively low cost (100 g cream or ointment = £7.90), it was likely to represent good value for NHS resource so long as it is used with caution and under careful supervision of a specialist in secondary care.

In thinking about the potential risks of prescribing potent, and in select cases very potent corticosteroids, the GDG considered it essential to build in monitoring to assess efficacy and adverse events. The time horizon of the economic model was too short (1 year) to explicitly consider annual monitoring in the long term; however, it is very likely that the extra cost of an annual GP or specialist visit would be offset by the avoidance of irreversible adverse events that are associated with inappropriate and unsafe use of corticosteroids.

The cost-effectiveness of topical treatments for children was not explicitly considered in the decision modelling undertaken for the guideline; however, the GDG considered the results broadly applicable to this population. They considered that once daily applications in children were likely to be more appropriate and that evidence of effectiveness for combination strategies are lacking. Therefore, they concluded that for children with mild to moderate psoriasis, once daily application of potent corticosteroids or vitamin D analogue were likely to represent the best value for NHS resource. They also considered how infrequent psoriasis occurs in children and that referral to secondary care may be justified.

M.4.3. Generalisability to other populations/settings

The analysis may be most applicable to patients with newly identified mild to moderate psoriasis, but the results may also be applicable to patients for whom topical therapy may be offered in addition to other therapies, such as phototherapy, systemic therapy or biologic therapy. These patients are likely to have much more widespread and/or severe disease and therefore topical therapy alone is likely to be insufficient and even inappropriate. However, the conclusion that topical corticosteroids offer good value for NHS resource and offer better value when combined with vitamin D analogue than TCF product is likely to apply to any population requiring topical therapies.

M.4.4. Comparisons with published studies

The findings from the NCGC original economic analysis are quite different from the results of the most similar published study by Bottomley and colleagues⁶². Bottomley and colleagues found 8 weeks of once daily TCF product to dominate other modelled strategies including once and twice daily vitamin D analogue followed by potent corticosteroid, potent corticosteroid followed by vitamin D analogue and 8 weeks of concurrent treatment with vitamin D analogue and potent corticosteroid. Although the analysis appears to have been executed well, the estimates of effect and resource use had limitations which called the conclusions of the analysis into question.

The biggest differences in the results of the NCGC analysis presented here and the analysis undertaken by Bottomley has to do with the treatment effect sizes used. In their analysis, concurrent treatment was found to be very ineffective, with just 14.9% of patients responding with a PASI75 compared to TCF product to which 50.3% of patients responded (RR=3.38). The NCGC analysis showed a much small difference between these treatments, with 65.1% of patients responding to concurrent treatment and 70.7% responding to TCF product (RR=1.09).

In addition, the estimate they used for quantity of topical used per 4-week treatment period was 92.6 g, compared to the estimate used in the NCGC analysis 134.0 g. Based on these estimates of resource use, the NCGC analysis assumes 4 weeks of TCF product costs £29.26 more than Bottomley and colleagues did. Furthermore, the difference between TCF product and concurrent treatment is different between the analyses. The additional cost of TCF product was £36.91 in Bottomley and more than twice that, £76.34, in the NCGC analysis. We performed a sensitivity analysis in which we assumed the same quantity of TCF product used by Bottomley and colleagues (i.e. 92.6 g, £61.27). The ICER for TCF product improved compared to the base case (£124,400 vs £174,545), but was still well above the NICE cost-effectiveness threshold of £20,000 per additional QALY.

The one thing that Bottomley and colleagues were able to capture that the NCGC analysis was not had to do with the potential disutilities associated with adverse events; however these inputs were not reported, were not included in their base case and, their impact on the results were not reported in full. The authors simply state that the influence of AEs ‘had no impact on the results.’

M.4.5. Conclusion

New economic analysis from a current UK NHS and PSS perspective comparing 118 different sequences of topical therapies found twice daily potent corticosteroids or concurrent treatment (morning/evening) with potent corticosteroid and vitamin D analogue to be the most cost-effective options for the first and second line treatment of patients with mild to moderate chronic plaque psoriasis. This conclusion was robust to the majority of sensitivity analyses undertaken.
- The base case and sensitivity analyses showed that the choice of third line treatment in a given sequence was highly uncertain. Depending upon the data used and assumptions made, third line treatment with twice daily coal tar, twice daily vitamin D analogue or once daily TCF product was likely to be most cost effective.

M.4.6. Implications for future research

Research into the longer term effectiveness and safety of available topical therapies would be valuable for future economic analyses undertaken in this area. In addition, it would be useful to identify the resource use associated with safe and effective methods of self-management with topicals, as there is quite a large degree of uncertainty about what ‘maintenance’ therapy actually means in the context of clinical practice.

Apart from any fair dealing for the purposes of research or private study, criticism or review, as permitted under the Copyright, Designs and Patents Act, 1988, no part of this publication may be reproduced, stored or transmitted in any form or by any means, without the prior written permission of the publisher or, in the case of reprographic reproduction, in accordance with the terms of licences issued by the Copyright Licensing Agency in the UK. Enquiries concerning reproduction outside the terms stated here should be sent to the publisher at the UK address printed on this page.

The use of registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant laws and regulations and therefore for general use.

The rights of National Clinical Guideline Centre to be identified as Author of this work have been asserted by them in accordance with the Copyright, Designs and Patents Act, 1988.

Bookshelf ID: NBK327696

Contents