Apremilast for treating active psoriatic arthritis

The committee heard from patient experts about the nature of psoriatic arthritis and their experiences of treatment. It heard that psoriatic arthritis is a lifelong condition that seriously affects people's quality of life. It can develop at a young age and affects all aspects of a person's life including education, work, self-care, and social and family life. The committee heard from the patient expert that skin symptoms can have a major psychological impact, and that joint symptoms can have an even greater impact on the psychological and functional aspects of living with the condition. The committee concluded that psoriatic arthritis substantially decreases quality of life.

The committee considered the current treatment pathway for people with psoriatic arthritis. It heard from clinical experts that after taking non-steroidal anti-inflammatory drugs and disease-modifying antirheumatic drugs (DMARDs) such as methotrexate, most people with non-responsive disease will have a tumour necrosis factor (TNF)‑alpha inhibitor, starting with the lowest cost drug as recommended in NICE technology appraisal guidance on etanercept, infliximab and adalimumab for the treatment of psoriatic arthritis and golimumab for the treatment of psoriatic arthritis. It heard from the clinical experts that use of more than 1 TNF‑alpha inhibitor is established practice in the NHS; if the disease fails to respond or loses response to the first TNF‑alpha inhibitor, or it causes adverse effects, a second TNF‑alpha inhibitor will often be used. The committee considered where apremilast would fit into this existing treatment pathway. It heard from the patient expert that when treatment with a TNF‑alpha inhibitor is contraindicated, or it is stopped because of loss of effectiveness or adverse effects (the clinical experts noted approximately 10% of patients per year stop TNF‑alpha inhibitor treatment), there may be no alternative treatments available. Therefore, patients and clinicians value having a range of treatment options available, and there is an unmet need for treatments that offer a different mechanism of action to the TNF‑alpha inhibitors or that are administered orally, as with apremilast (a PDE4 inhibitor).

The committee was aware that apremilast had the same marketing authorisation as the currently recommended biological treatments, but that the company had stated that apremilast would be used before these treatments in clinical practice, based on its oral route of administration, safety profile compared with current biological and conventional DMARD treatments, no specific requirements in the marketing authorisation for regular monitoring, and a cheaper cost compared with current biological therapies. The committee was also aware of a written statement from the clinical expert that apremilast could be considered an alternative first- or second-line drug, because it was likely more effective than methotrexate. However, the written statement from the clinician had noted that placement in the pathway would also depend on treatment cost. The committee heard from the clinical experts that it would be useful to have an additional treatment option before TNF‑alpha inhibitors, because the psoriatic arthritis population is heterogeneous and some people cannot tolerate DMARD therapy, or their disease does not respond adequately to it. The committee concluded that it was possible that apremilast could be used as a treatment before TNF‑alpha inhibitors, but that any use or positioning of apremilast would need to be supported by clinical and cost‑effectiveness evidence, particularly because several effective treatment options are already recommended for psoriatic arthritis.

The committee considered the most appropriate comparators for this appraisal. It was aware that in June 2015, NICE published guidance on ustekinumab for treating active psoriatic arthritis which, as an IL12/23 inhibitor, offered a different mechanism of action to the TNF‑alpha inhibitors. However, it accepted that current usage of this drug was likely to be low, both because it had only relatively recently received a positive recommendation, and also because the recommendation is more restrictive than the currently recommended TNF‑alpha inhibitors (ustekinumab is recommended as a treatment option only if treatment with TNF‑alpha inhibitors is contraindicated but would otherwise be considered, or if the person has had treatment with 1 or more TNF‑alpha inhibitors). The committee was also aware that certolizumab pegol (another TNF‑alpha inhibitor) is another possible treatment option for people with psoriatic arthritis; however, it heard from the clinical experts that it is rarely used in clinical practice. The committee concluded that the most appropriate comparators for this appraisal were the TNF‑alpha inhibitors adalimumab, etanercept, infliximab and golimumab (because they have a similar marketing authorisation to apremilast, and are the most commonly used treatments in clinical practice after the failure of a DMARD) and that ustekinumab could be considered as a comparator if it became relevant to consider making a recommendation specifically for a population for whom TNF‑alpha inhibitors are not appropriate.

The committee heard from the clinical and patient experts that although methotrexate works well, some people fear the adverse effects associated with it (such as hair loss, nausea and lethargy) and the need for frequent blood tests. The experts stated that apremilast may be better tolerated, although it is associated with a higher incidence of diarrhoea initially compared with some DMARDs such as leflunomide. The clinical experts stated that there is no evidence on whether apremilast is better tolerated than TNF‑alpha inhibitors and that, in general, the TNF‑alpha inhibitors are well tolerated; apremilast is no better or worse than the TNF‑alpha inhibitors, and most patients do not experience unacceptable problems. The clinical experts also suggested that, as with any new treatment, apremilast would need extra monitoring because its long-term adverse events are unknown. The committee was aware of new evidence about the adverse effects of apremilast that the company had submitted in response to the appraisal consultation document, which provided further evidence about the adverse event profile for apremilast. The committee concluded that apremilast has an acceptable adverse event profile in people with active psoriatic arthritis.

The committee considered the evidence presented by the company on the clinical effectiveness of apremilast. It noted that the main sources of evidence were the PSA-002, PSA-003 and PSA-004 trials that compared apremilast (20 mg and 30 mg) with placebo in patients with active psoriatic arthritis (3 or more swollen and tender joints for at least 6 months) that had not responded to treatment with up to 3 DMARDs or 1 TNF‑alpha inhibitor. The committee noted that the trials were well conducted and showed that apremilast is more effective than placebo after 16 weeks of treatment for a number of joint, skin and soft tissue outcomes; the primary outcome was ACR20, with a response experienced by 37% of people having apremilast compared with 19% having placebo (p≤0.0001). The clinical experts noted that apremilast was associated with a similar ACR20 response to methotrexate. The committee acknowledged that in response to the appraisal consultation document the company stated that it considered this opinion to be subjective, because little comparative evidence is available in this area. The committee also noted that apremilast was effective for associated problems such as dactylitis and enthesitis. The committee agreed that apremilast was a clinically effective treatment compared with placebo.

The committee considered the more stringent ACR outcomes (ACR50 and ACR70) presented in the apremilast trials. It heard from the clinical experts that although ACR20 is an accepted outcome measure for treatments of psoriatic arthritis and was the primary outcome in the apremilast trials, people may still have painful and swollen joints and that people start to notice a benefit at ARC50 or ACR70. The committee agreed that there was a difference between apremilast and placebo but that the absolute differences were less than those seen for ACR20.

The committee considered the evidence from the company's network meta-analysis that compared apremilast with TNF‑alpha inhibitors in the total population, and in the population who had not been treated with TNF‑alpha inhibitors. The committee heard from the evidence review group (ERG) that the methods used to identify both published and unpublished studies for the network meta-analysis were appropriate, and the studies were mostly well reported. The committee discussed the ERG's concerns that the placebo responses for some outcomes were high which made it difficult to compare the relative efficacies of apremilast with the different comparators. The committee noted that the results showed that apremilast had a clinical benefit compared with placebo. However, apremilast demonstrated less clinical benefit than any of the TNF‑alpha inhibitors, in either population. The committee concluded that apremilast is not as clinically effective as the TNF‑alpha inhibitors for treating psoriatic arthritis.

The committee considered the HAQ‑DI outcome used by the company to calculate functional capacity and to assess disease progression. It heard from the ERG that there were uncertainties about the results from the apremilast trials because they were not blinded after 24 weeks and there were no stopping rules, which was likely to have influenced the HAQ‑DI results. The committee noted that the company had provided evidence to argue against this in its response to the appraisal consultation document; for example, the company stated that participants remained blinded to initial treatment and dose during the unblinded period. However, the committee remained concerned that, in comparison with more objective measures of disease progression such as radiographic assessments, there was a higher possibility of bias.

The committee considered the lack of radiographic assessment in the apremilast trials. It heard from the clinical experts that it would be difficult to justify using apremilast early in the treatment pathway (before TNF‑alpha inhibitors) without evidence that it can prevent radiological progression, because there is evidence to show that TNF‑alpha inhibitors slow disease progression. The committee also heard from the patient experts that they want treatments that can stop the disease from progressing. It noted that the company had stated in its response to the appraisal consultation document that the relationship between radiographic progression and functional capacity was unclear, and that other measures such as disease activity were equally, if not more, important when considering the impact of disease on quality of life. The committee accepted that it may be necessary to interpret radiographic evidence with caution, and that disease activity outcomes play an important role in functional capacity. However, it noted that apremilast not only lacked radiographic evidence about disease progression, but had consistently shown the worst performance of any active comparator for all outcomes presented in the network meta-analyses. Because it is a new treatment, there is a lack of long-term clinical effectiveness data for apremilast. The committee concluded that the lack of radiographic evidence and the clinical-effectiveness evidence did not support the use of apremilast before TNF‑alpha inhibitors in clinical practice.

The committee considered the company's revised model which, as in the original base case, compared treatment sequences with and without apremilast, rather than comparing apremilast with a single comparator. This provided a revised base-case incremental cost‑effectiveness ratio (ICER) of approximately £19,500 per quality-adjusted life year (QALY) gained when adding apremilast to a treatment sequence of adalimumab, etanercept, and best supportive care. Apremilast remained cost effective (when assuming a maximum acceptable ICER of £30,000 per QALY gained) in exploratory analyses, including when varying apremilast HAQ-DI progression in relation to best supportive care (£22,700 to £29,100 per QALY gained). The committee accepted that the use of treatment sequences was a valid approach to modelling.

The committee considered whether the structural and parameter assumptions in the company's treatment sequences in the revised base case reflected clinical practice. It noted that most analyses by the company compared treatment sequences that had a different number of active comparators before progression to best supportive care, with the base case comparing 3 active treatments for the apremilast group with 2 for the comparator group. The committee agreed that, in clinical practice, patients would likely receive more than the 2 active treatments patients were assumed to receive in the comparator group before they progressed to best supportive care. This was because there are a number of active comparators available for treating psoriatic arthritis, particularly since the positive recommendation for ustekinumab. The committee also considered that models comparing sequences, rather than more traditional direct comparisons, created additional uncertainty in the model. Treatment sequences of different lengths may exacerbate uncertainties in the model, which may also be less easily identifiable, because they are less likely to affect each arm equally than with direct comparisons or equal length sequences. The committee further understood from the assessment group analyses that, assuming all other things were equal, replacing apremilast in the intervention group of the company revised base case with any of the TNF‑alpha inhibitors would result in a QALY gain over the comparator sequence. The committee concluded that in order to prevent the model being confounded by any QALY gain occurring only because of one group in the model having an additional active treatment, in a selected and unrealistically short sequence, it was more informative to make inferences from modelling the same number of active comparators in each treatment sequence.

The committee noted that the company had presented a limited exploratory analysis using treatment sequences of equal length in which apremilast was used instead of adalimumab in a sequence of adalimumab, etanercept, golimumab and best supportive care. However, the committee noted that this needed to be seen in the context of the ERG's multiple calculations using sequences with an equal number of active comparators, and also noted that the company considered this scenario to be of limited relevance. The committee also noted that the analyses should be consistent with the direct clinical and cost differences between the TNF‑alpha inhibitors and apremilast.

The committee considered the company's assumptions about the improvement and progression of joint symptoms (measured using HAQ‑DI). It noted that these were key drivers of the economic model and that people whose disease continued to respond to treatment at the end of the trial period retained the same HAQ‑DI score (that is, apremilast was assumed to halt HAQ‑DI progression while people remained on treatment, therefore zero HAQ‑DI progression was applied). The committee noted that the company's rationale for assuming that apremilast halts disease progression was based on acceptance in previous NICE appraisals for psoriatic arthritis that TNF‑alpha inhibitors halt disease progression. The committee was aware that the assumption that TNF‑alpha inhibitors halt disease progression was supported radiographically and also by clinical practice evidence over a number of years. However, there was uncertainty about whether this assumption was equally relevant for apremilast, which has a different mechanism of action and limited evidence of use in clinical practice because it is a relatively new treatment. The committee also noted that people who progressed to best supportive care were assumed to experience subsequent natural progression of their disease, resulting in an increase (worsening) in HAQ‑DI score over time of 0.006 every 28 days, up to a maximum score of 3. The committee noted that this score appeared high but heard from the clinical experts that, although it is not possible to know if people would experience a linear progression of disease, the clinical experts considered that the increase in HAQ‑DI over time is likely to be within the same range as that used by the company. The committee heard from the ERG that experience with rheumatoid arthritis shows that HAQ‑DI does not have a linear trajectory; the rate of progression of the disease slows down over time. However, the committee also noted comments from the company in response to the appraisal consultation document that the linearity of HAQ‑DI progression was hypothetical and that NICE's technology appraisal guidance on ustekinumab for treating active psoriatic arthritis had assumed linear progression. The committee also noted that patients with the best HAQ‑DI responses would be likely to remain in the trials, making the HAQ‑DI appear to improve over time. The committee acknowledged that there is a lack of evidence to inform these model assumptions, and this added uncertainty to the model. However, the assumption that apremilast completely halts HAQ‑DI progression represented a best-case scenario that was not supported by clinical evidence (see sections 4.8, 4.9 and 4.10).

The committee considered the use of HAQ‑DI and PASI scores mapped to EQ‑5D to produce utility values of health in the company's original base case. The committee noted that the utility values in the company's revised base case were derived from the apremilast trial. Although this reflected the preferences of the committee as expressed in the appraisal consultation document, the committee noted that this had little impact on results compared with the values used in the original base case. The committee was also surprised at the estimates of utility, which appeared very low and similar to technologies for end of life conditions. However, the committee agreed that the company had used a legitimate source for utility values by using the available trial data, and accepted the utility values for its decision-making.

The committee discussed the costs included in the model, particularly the monitoring costs for apremilast treatment. It noted that in response to the appraisal consultation document the company had stated that monitoring costs for apremilast should not be included because there were no specific requirements for screening or regular monitoring, but that it had updated its revised base case to include an equal level of monitoring for all active treatments. The committee heard from the clinical experts that, as with any new drug, apremilast would initially need more monitoring compared with the current standard of care. It therefore concluded that the revised model had correctly accounted for monitoring costs for apremilast.

The committee considered the assumption of different trial periods for apremilast (16 weeks) and TNF‑alpha inhibitors (12 weeks) for PsARC responses. The committee heard from the ERG that the use of different time points could favour apremilast and that, if the trial period for TNF‑alpha inhibitors were also increased to 16 weeks, the PsARC responses may increase. The clinical experts agreed that using different trial periods could influence the results. The committee acknowledged that the company had carried out a scenario analysis altering the length of the apremilast trial period to 24 weeks but leaving the TNF‑alpha inhibitor response at 12 weeks. The committee concluded that the longer trial period of apremilast could have given a relatively optimistic case for apremilast compared with other comparators.

The committee considered the company's assumptions for placebo responses in the original and revised model. It noted that in the original model, the placebo response rate was discounted from best supportive care, but not from the absolute response rates of apremilast or the TNF‑alpha inhibitors used in the model. However, in the revised base case, the company had included a placebo response for best supportive care. The committee agreed that inclusion of placebo response rates in the model was necessary and accepted this revision to the model.

The committee noted that the company's original base case results were based on uncertain assumptions. It appreciated that the company had attempted to address this uncertainty by making several changes in its revised model (including equal levels of monitoring for apremilast and TNF‑alpha inhibitors, a placebo response for best supportive care, and utility values derived from the apremilast trial), and also by presenting several exploratory analyses. However, most ICERs presented by the company were based on treatment sequences with an unequal number of treatments, which was not the committee's preference (see sections 4.11 and 4.19). The committee therefore went on to consider the exploratory analyses presented by the ERG. The committee noted that the ERG had based its analyses on the revised company base case and, therefore, as in the company revised base case, it accounted for several uncertainties in the original base case. Also, the ERG had used the committee's preferred treatment sequences, with an equal number of active comparators before progression to best supportive care, for its exploratory analyses. The committee concluded that the exploratory analyses presented by the ERG were the most appropriate for decision-making.

The committee considered the results for apremilast as a treatment before TNF‑alpha inhibitor therapy, using its preferred exploratory analyses from the ERG (see sections 4.11 and 4.17). The committee noted that all the ERG's sequences in which apremilast was the first treatment in a sequence (after DMARDs) resulted in cost savings but also a QALY loss, resulting in ICERs that reflected 'savings per QALY lost'. For example, when comparing a sequence of apremilast, adalimumab, etanercept and best supportive care with adalimumab, etanercept, golimumab, and best supportive care, and when using the committees preferred assumption of some HAQ-DI progression for apremilast (at half the rate of that for best supportive care) there was a cost saving of £6,739 in the apremilast sequence, but a QALY loss of −0.368, resulting in an ICER of £18,300 saved per QALY lost. The committee considered this to be the most plausible scenario because it used its preferred assumptions, and also because the results were consistent with the clinical and cost data; that is, when compared with TNF‑alpha inhibitors, apremilast cost less but was also the least effective active treatment. The committee noted that, in situations in which an ICER is derived from a technology that is less effective and less costly than its comparator, the commonly assumed decision rule of accepting ICERs below a given threshold is reversed, and so the higher the ICER, the more cost effective a treatment becomes. The committee was aware that psoriatic arthritis is a chronic and progressive condition, that patients want treatments that stop disease progression (see section 4.10), and that apremilast was the least effective treatment in the company analyses. Taking all of the above into account, the committee agreed that the ICER for apremilast was not high enough to compensate for the clinical effectiveness that would be lost. It therefore concluded that apremilast was not a cost‑effective option compared with TNF‑alpha inhibitors for people with psoriatic arthritis that has responded inadequately to DMARDs.

The committee considered whether there was any evidence to consider apremilast as a treatment after TNF‑alpha inhibitor therapy, or for people who could not take TNF‑alpha inhibitors. It noted that evidence in this area was limited. The available clinical effectiveness evidence for apremilast was mostly for a population who had not previously had TNF‑alpha inhibitors. The cost‑effectiveness evidence was limited because the company had rejected this possible positioning of apremilast, even though such comparisons (particularly with ustekinumab) were listed in the final scope issued by NICE. The company had presented 2 direct comparisons of apremilast with best supportive care, and when assuming apremilast HAQ‑DI progression at a rate half that of best supportive care, the ICER for apremilast was £21,700 per QALY gained. The committee noted, however, that the company had not explored the analyses further because it did not consider best supportive care to be an appropriate comparator. Following the publication of NICE's technology appraisal guidance on ustekinumab for treating active psoriatic arthritis, and given the range of other treatments available for psoriatic arthritis, there are a number of other possible treatments used after TNF‑alpha inhibitors that would be available before best supportive care, and these had not been explored as comparators. The committee also considered the ERG's scenarios for apremilast used after TNF‑alpha inhibitors, which included the committee's preferred model assumption of the same number of active treatments in each sequence. The committee was aware of the ERG's comments about the validity of its exploratory analyses and agreed that as these were the only scenarios presented for apremilast used after TNF‑alpha inhibitors, they should be taken into account in its decision-making. The committee noted that in all the ERG's exploratory analyses the apremilast treatment sequence resulted in cost savings but a QALY loss, resulting in ICERs that reflected 'savings per QALY lost'. For example, a treatment sequence in which apremilast replaced golimumab in a sequence of adalimumab, etanercept, golimumab and best supportive care, assuming HAQ-DI progression at a rate equal to half of best supportive care, resulted in a cost saving of £5,343 and a QALY loss of -0.362, with an ICER of £14,800 saved per QALY lost. The committee agreed that this was the most plausible scenario that had been presented because it used the committee's preferred assumptions about treatment sequences with an equal number of treatments and some HAQ-DI progression for apremilast, the results were consistent with the clinical and cost data (that is, when compared with TNF‑alpha inhibitors, apremilast cost less but was also the least effective active treatment), and also because of the limited evidence presented by the company. The committee agreed that the ICER for apremilast was not high enough to compensate for the clinical effectiveness that would be lost. It therefore concluded that apremilast could not be recommended as a treatment after TNF‑alpha inhibitors. It was unable to make recommendations for its use when people cannot take TNF‑alpha inhibitors, because of a lack of evidence for its use in these circumstances.

The committee discussed whether apremilast is considered innovative. It heard from clinical and patient experts that apremilast may provide an additional treatment option for patients, because of its different mode of action and oral formulation. However, given its conclusion on clinical efficacy (see sections 4.6 to 4.8) the committee considered that apremilast was not a step-change in treatment. The committee concluded that there were no additional gains in health-related quality of life over those already included in the QALY calculations, and that there was no need to change its conclusions on that basis.

The committee was aware of NICE's position statement on the Pharmaceutical Price Regulation Scheme (PPRS) 2014, and in particular the PPRS payment mechanism, and accepted the conclusion 'that the 2014 PPRS payment mechanism should not, as a matter of course, be regarded as a relevant consideration in its assessment of the cost effectiveness of branded medicines'. The committee heard nothing to suggest that there is any basis for taking a different view with regard to the relevance of the PPRS to this appraisal. It therefore concluded that the PPRS payment mechanism was not relevant for its consideration of the cost effectiveness of any of the technologies in this appraisal.