Alitretinoin for the treatment of severe chronic hand eczema

The manufacturer approached the decision problem by comparing alitretinoin with ciclosporin, oral and topical PUVA (psoralen and long-wave ultraviolet radiation), and azathioprine. The population considered was adults with severe chronic hand eczema that is unresponsive to potent topical corticosteroids. The primary outcome measures outlined in the decision problem were overall severity of chronic hand eczema (as defined by the PGA), modified total lesion symptom score (mTLSS), patient's global assessment of improvement, time to response, time to relapse and a disease-specific quality of life measure, namely the dermatology life quality index (DLQI).

The primary outcome measure used in the clinical trials was severity of chronic hand eczema as defined by the PGA. This combined the grading of disease severity against a photographic guide with an indication of symptoms (pruritus and/or pain) and degree of functional impairment. The PGA describes five severity states for chronic hand eczema (clear, almost clear, mild, moderate and severe), and a combined 'clear or almost clear' category was used to define response to treatment in the trials.

The manufacturer's submission presented evidence on the clinical effectiveness of alitretinoin from two multinational randomised controlled trials (RCTs): BAP0003, a 12-week phase II trial (n = 319) comparing three doses of alitretinoin (10 mg, 20 mg and 40 mg daily) with placebo; and BAP00089, a 24-week phase III trial (n = 1032) evaluating daily 10 mg and 30 mg doses of alitretinoin versus placebo. It also presented evidence from BAP00091, an extension of the BAP00089 RCT in which non-responding and responding–relapsing people were followed up for 24 weeks. In BAP00091, all people (n = 360) in BAP00089 whose eczema had not responded or who had disease relapse within 24 weeks of treatment received a further 12-week or 24-week course of either 10 mg or 30 mg of alitretinoin or placebo (people from BAP00089 who had received placebo were assigned to receive placebo again; people who had received alitretinoin were given a further course of treatment with the same dose of alitretinoin or assigned to receive placebo). All trials included people whose eczema had not responded to potent topical corticosteroids. The BAP00089 RCT included people with 'severe' eczema as defined by PGA score. The BAP0003 trial included people with either 'moderate' or 'severe' eczema as defined by PGA score.

Both RCTs found that alitretinoin treatment resulted in a greater proportion of people with hands clear or almost clear at 12 and 24 weeks compared with placebo, as assessed by PGA score and patient's global assessment of improvement. The differences were statistically significant (although in the BAP0003 trial, only the 40 mg dose of alitretinoin gave statistically significant results compared with placebo). In the BAP00089 trial, 47.7% of people were reported as having clear or almost clear skin by week 24 of treatment with 30 mg of alitretinoin, compared with 16.6% for placebo (p < 0.001). The BAP00089 trial also measured rates of remission and found that among people whose eczema had responded to alitretinoin treatment, 30% treated with 30 mg and 37% treated with 10 mg relapsed during the 24-week follow-up period. The manufacturer reported that in the BAP0003 study, 26% of people whose eczema had responded to treatment with alitretinoin relapsed (mTLSS score of 75% of the baseline value) within 12 weeks of the end of the treatment.

In the extension study (BAP00091), participants were divided into two cohorts. Cohort A consisted of 117 people whose eczema had relapsed within 24 weeks of treatment, and a double-blind design was used. People were assigned to receive the same dose of alitretinoin as in BAP00089 or placebo; those who had received placebo in BAP00089 were again assigned to the placebo group. In this trial, 21 people were given 10 mg of alitretinoin, 49 people were given 30 mg of alitretinoin and 47 people were given placebo, for a period of 12 or 24 weeks. A statistically significantly greater proportion of people treated again with 30 mg of alitretinoin had a PGA state of hands clear or almost clear than those treated with placebo (79.6% and 8.3% respectively, p < 0.001). Cohort B consisted of 243 people whose eczema had not responded to treatment in the original RCT. All were given 30 mg of alitretinoin and an open-label design was used. Nearly 50% of people whose eczema had not initially responded to treatment after 24 weeks responded to a further 12-week or 24-week course of 30 mg of alitretinoin once daily.

The manufacturer also provided details of subgroup analyses from the BAP00089 trial. The 30 mg dose of alitretinoin resulted in a higher proportion of people with hands clear or almost clear than placebo in people with hyperkeratotic disease (54% versus 12%), pompholyx disease (33% versus 30%), and hyperkeratotic and pompholyx disease together (33% versus 12%). It was not stated whether these differences were statistically significant.

The manufacturer reported that information on health-related quality of life (HRQoL) was collected only during the phase II BAP0003 study and that 51.4% of people in both treatment groups (alitretinoin and placebo) completed DLQI questionnaires. The median change in HRQoL from baseline was greater with alitretinoin than with placebo (–3 [for doses of 20 and 40 mg of alitretinoin] and –2 respectively). The findings were not statistically significant, but the manufacturer pointed out that this may have been because of the lack of statistical power of the study. The manufacturer did not include the DLQI or any other measure of HRQoL in any subsequent trials or analyses.

The primary source of data on adverse events in the manufacturer's submission was the phase III RCT (BAP00089). Treatment-related serious adverse events with alitretinoin were rare (an incidence of 1% at a dose of 30 mg). The most common adverse event was headache (20% at 30 mg; 11% at 10 mg), and a small proportion of people had elevated blood triglycerides (3% at 30 mg; 1% at 10 mg) and high total cholesterol (14% at 30 mg; 3% at 10 mg). The number of people who withdrew from the trial because of adverse events was 39 (9.5%) for 30 mg of alitretinoin and 24 (5.7%) for 10 mg of alitretinoin. The number of people who refused to continue treatment for other reasons was 16 (3.9%) for 30 mg of alitretinoin and 24 (5.7%) for 10 mg of alitretinoin.

The manufacturer pointed out that there were no trials that compared alitretinoin directly with any of the comparators specified in the scope for the appraisal. It explained that subsequent searches were carried out to identify trials that assessed the efficacy of PUVA, ciclosporin and azathioprine for the treatment of chronic hand eczema. This search identified 13 trials of PUVA for the treatment of chronic hand eczema, of which eight met the criteria for inclusion in the review. One trial of ciclosporin and no trials of azathioprine were identified. The manufacturer explained that a mixed-treatment comparison could not be carried out because none of the RCTs using PUVA or ciclosporin had a placebo control arm, and therefore no common link could be established between the trials of alitretinoin, PUVA and ciclosporin.

The manufacturer submitted a cost-effectiveness analysis from a de novo Markov-based patient-level model using a hypothetical cohort of people with severe chronic hand eczema. The demographic characteristics of the model population reflected those of the participants in the BAP00089 trial, and 15% of the women were assumed to be of childbearing potential. The model had five health states that were defined according to PGA score: severe, moderate and mild chronic hand eczema, remission (people whose chronic hand eczema was rated as 'clear' or 'almost clear' by 24 weeks), and refractory disease (people whose chronic hand eczema was rated 'moderate', 'mild' or had returned to a PGA state of 'severe' at 24 weeks). The model was designed to compare oral alitretinoin with PUVA, ciclosporin, azathioprine and best supportive care. The model had a 3-year time horizon, and a treatment course of alitretinoin was assumed to be given for between 12 and 24 weeks at an initial dosage of 30 mg once daily.

The efficacy estimates for alitretinoin in the model were taken from the phase III clinical trial (BAP00089) for the first treatment cycle, and from cohort A of the phase III extension trial (BAP00091) for subsequent treatment cycles. Estimates of the efficacy of the comparators were obtained from a panel of seven dermatologists. Data on the number of adverse events and the probabilities of dose reduction or withdrawal from treatment were informed by BAP00089 or by the manufacturer's assumptions. Time to relapse following remission was informed by the BAP00089 trial for alitretinoin and by expert clinical opinion for the comparators. For alitretinoin, the estimates of the proportion of people who move to each PGA state after initial treatment were obtained from the BAP00089 trial, and retreatment estimates were obtained from the BAP00091 trial. The corresponding estimates for the comparator interventions were obtained from expert opinion.

The utility values for all health states were derived using data collected from the BAP0003 trial to predict DLQI scores that correspond to each PGA state. A published algorithm of the relationship between DLQI scores and EQ-5D scores in people with psoriasis was then used to predict EQ5D-based utility values from DLQI scores. The model applied the utility scores associated with the 'severe' PGA state to people whose disease was rated as severe and who were still receiving treatment, and to people whose disease was deemed to be refractory. The 'moderate' and 'mild' utility scores were applied to people receiving treatment whose disease was rated moderate and mild, respectively, on the PGA scale. The utility scores for the states of 'clear' and 'almost clear' were averaged to provide a single utility score that was applied to people whose disease was in remission. The manufacturer also provided a set of alternative utility estimates from an unpublished study by Augustin (Augustin M: unpublished data 2008). These EQ-5D scores were predicted from the observed average DLQI scores of the people within each PGA state. Adverse events were assumed to have no impact on HRQoL.

It was assumed that if an adverse event occurred (either headache or hyperlipidaemia), 20% of people with headache and 40% of people with hyperlipidaemia would switch to a lower dose (10 mg of alitretinoin, once daily); treatment would continue unchanged at 30 mg of alitretinoin for the remainder of people with headache or hyperlipidaemia. It was then assumed that those people who switched to the lower dose and who experienced a subsequent adverse event had a 20% probability of withdrawal owing to headache and a 40% probability of withdrawal owing to hyperlipidaemia. The people who withdrew would enter the refractory state, and the remaining people in this group would continue treatment with 10 mg of alitretinoin. The costs associated with treatment, monitoring and adverse events were included in the model.

The manufacturer's original base case resulted in an incremental cost-effectiveness ratio (ICER) of £8614 per quality-adjusted life year (QALY) gained for alitretinoin compared with ciclosporin. Alitretinoin dominated PUVA. A comparison of alitretinoin with azathioprine resulted in an ICER of £10,612 per QALY gained.

The manufacturer carried out two subgroup analyses. The first subgroup was people with hyperkeratotic disease. For this subgroup, the manufacturer adjusted the efficacy data for alitretinoin to reflect the improved efficacy that had been observed in the BAP00089 trial predominantly in people with hyperkeratotic disease. The second subgroup analysis was in women of childbearing potential. The efficacy was assumed to be the same in these women as in the base case, but the care of these women was assumed to incur additional costs associated with conception counselling and pregnancy testing. The consequences of the Pregnancy Prevention Programme not working were not considered.

The manufacturer's subgroup analyses for people with hyperkeratotic disease resulted in an ICER of £11,177 per QALY gained for alitretinoin compared with ciclosporin. Alitretinoin dominated PUVA. The comparison of alitretinoin with azathioprine resulted in an ICER of £13,174 per QALY gained. The manufacturer's subgroup analyses for women of childbearing age resulted in ICERs for alitretinoin of £9109, £11,038 and £54 per QALY gained, compared with ciclosporin, azathioprine and PUVA respectively.

In response to a request for clarification from the ERG, the manufacturer submitted a revised model comparing alitretinoin with best supportive care. The manufacturer's revised base case resulted in an ICER of £12,931 per QALY gained. The ICER was £15,018 per QALY gained for people with hyperkeratotic disease and £26,013 per QALY gained for people with hyperkeratotic and pompholyx disease.

The manufacturer undertook a one-way sensitivity analysis of the time horizon of the revised model. Using just a 1-year (rather than a 3-year) time horizon resulted in an ICER of £21,562 per QALY gained.

The ERG highlighted a number of concerns with the clinical and cost effectiveness information in the manufacturer's submission, including:

• the validity of the efficacy estimates for the comparators

• the possibility that the population and some assumptions in the model may not reflect clinical practice in England and Wales

• the high degree of uncertainty because derived utility values were used rather than directly observed HRQoL values

• errors in the model's visual basic for applications (VBA) code.

The ERG regarded the comparisons of alitretinoin with azathioprine, ciclosporin and PUVA in the original manufacturer's submission to be of limited value. This was because the efficacy data for the comparators were based on expert clinical opinion only. Although the ERG accepted that there was no appropriate clinical trial evidence, it did not think the elicitation process used was sufficiently rigorous. It therefore questioned the validity of the efficacy estimates for the comparators used in the model, and noted the absence of any quantification of the uncertainty around these estimates. The ERG therefore viewed the comparison of alitretinoin with placebo in the revised model to be of greater relevance, and focused its evaluation on this aspect of the model.

The ERG questioned whether the model population (people with severe chronic hand eczema as defined by PGA score) reflected the population of people with corticosteroid-refractory chronic hand eczema for whom clinicians would aim to provide treatment.

The ERG was unsure of the validity of some of the model assumptions. These included the assumptions that people would stop treatment as soon as their disease responded, even if this was after only 4 or 8 weeks of treatment; that all people who relapse return to the PGA severe state, even though the time to relapse was informed by trial data that used a definition of relapse based on return to 75% baseline mTLSS; and that people receiving alitretinoin would visit a dermatologist every 4 weeks.

The ERG viewed the derived utility values used in the model as a major source of uncertainty for the cost-effectiveness analysis. It also considered that the utility estimates obtained using the directly observed relationship between PGA state and DLQI score from the Augustin study may be a more appropriate basis for modelling than the analysis of change in DLQI score calculated based on PGA state from the BAP0003 trial.

The ERG stated that there were errors in the model's VBA code. This meant that the first 4 weeks of every treatment cycle except the first cycle were omitted from the model. It also pointed out that adverse events associated with alitretinoin had been removed from the revised model that compared alitretinoin with best supportive care.

The ERG carried out an additional exploratory cost-effectiveness analysis using the manufacturer's original model. It explained that the results given by the manufacturer were not fully incremental, consisting of pairwise comparisons between alitretinoin and the comparators. The ERG explained that integrating the supportive care arm from the revised model into a fully incremental analysis was possible because the manufacturer removed adverse events from the revised model and did not report on the adverse-event profile associated with supportive care. The ERG's incremental analysis found that alitretinoin extendedly dominated ciclosporin, alitretinoin dominated PUVA, best supportive care dominated azathioprine, and the comparison of alitretinoin with best supportive care resulted in an ICER of £12,931 per QALY gained.

The ERG also conducted exploratory sensitivity analyses using the manufacturer's revised model (which compared alitretinoin with best supportive care). Using the utility estimates from the Augustin study and the assumption that people (except women of childbearing potential) see a dermatologist once every 6 weeks if they are taking alitretinoin and once every 12 weeks if they are receiving best supportive care (rather than once every 4 weeks) resulted in an ICER of £27,997 per QALY gained for alitretinoin compared with best supportive care.

Using the ERG-modified VBA code so that people with relapsing disease moved to the appropriate PGA state (30.6% of people whose disease relapsed moved to the moderate state and the remainder to the severe state) resulted in an ICER of £29,864 per QALY gained for alitretinoin compared with best supportive care. Reinstating adverse events for alitretinoin resulted in an ICER of £29,200 per QALY gained for alitretinoin compared with best supportive care.

Using all modifications described in sections 3.26 and 3.27, but keeping the utility data from the original model (taken from BAP0003), resulted in an ICER of £15,084 per QALY gained for alitretinoin compared with best supportive care. Using all modifications described in 3.26 and 3.27, including the alternative utility data from the Augustin study, resulted in an ICER of £30,918 per QALY gained for alitretinoin compared with best supportive care.