Psoriatic arthritis - etanercept and infliximab

4. Evidence and interpretation

The Appraisal Committee (appendix A) considered evidence from a number of sources (appendix B).

4.1. Clinical effectiveness

4.1.1. Etanercept

4.1.1.1. Two randomised controlled trials (RCTs, n = 265) were included in the Assessment Report. These trials were double-blind and placebo-controlled and were deemed to be of good quality by the Assessment Group. Neither trial required there to have been a failed response to DMARDs. In both RCTs, participants were allowed to have concomitant corticosteroids and NSAIDs. In addition, people were allowed DMARD monotherapy with methotrexate only. The controlled phase of both trials was followed by an open-label follow-up period during which the active drug was administered to all participants. The primary outcome in the 12-week smaller trial (n = 60) was PsARC response, whereas in the 24-week larger trial (n = 205) it was ACR20.

4.1.1.2. Etanercept was statistically significantly more effective than placebo for all arthritis outcomes in both RCTs, with the exception of the ACR70 in the smaller trial. In the subgroups of participants who also had at least 3% of their body surface area affected by psoriasis, the effect of etanercept was statistically significantly greater than placebo as measured by the PASI 75 in both studies but not by the PASI 25 and PASI 50 at 12 weeks in the smaller RCT, and PASI 90 at 24 weeks in the larger RCT. The pooled results at 12 weeks showed a statistically significant difference in favour of etanercept in all outcomes apart from PASI 75, and no statistical heterogeneity was detected: PsARC relative risk [RR] 2.6 (95% confidence interval [CI], 2.0 to 3.5); ACR50 RR 10.8 (95% CI, 4.5 to 26.3); PASI 75 RR 2.3 (95% CI, 0.9 to 6.0); PASI 50 RR 2.4 (95% CI, 1.4 to 4.2); HAQ 49% change from baseline (95% CI, 38.5 to 59.4). The 24-week results in the larger RCT were: PsARC RR 3.0 (95% CI, 2.1 to 4.4); ACR50 RR 9.5 (95% CI, 3.5 to 25.8); PASI 75 RR 7.0 (95% CI, 1.7 to 29.6); HAQ 47% change from baseline (95% CI, 32.5 to 61.9). In the larger RCT, the Total Sharp Score radiographic assessment that measures annualised rate of progression at 24 weeks was statistically significantly lower in the group that had received etanercept than in the group that had received placebo.

4.1.1.3. In the larger RCT, subgroup analysis by concomitant methotrexate treatment (stratified at randomisation) indicated that the effect of etanercept did not depend on the concomitant use of methotrexate. The uncontrolled follow-up data reflected those in the controlled period and the 1-year Total Sharp Score data indicated that, on average, there was no clinically significant disease progression.

4.1.1.4. In the larger RCT, at 24 weeks 65 of the 101 participants receiving etanercept experienced an adverse event (64%) compared with 69 of the 104 participants in the placebo group (66%) (RR 1.0; 95% CI, 0.8 to 1.2). Four people in each group experienced serious adverse events (RR 1.0; 95% CI, 0.3 to 4.0) and one person in each group withdrew because of adverse events (RR 1.0; 95% CI, 0.1 to 16.2).

4.1.2. Infliximab

4.1.2.1. One RCT (the infliximab multinational psoriatic arthritis controlled trial, or IMPACT) was included in the Assessment Report and deemed to be of good quality. All 104 participants had to have had PsA that had failed to respond to at least one DMARD. People were allowed to have concomitant corticosteroids and NSAIDs plus DMARD monotherapy with any agent. The controlled phase of the RCT was followed by an open-label follow-up period during which the active drug was administered to all participants.

4.1.2.2. The primary outcome measure in the RCT was ACR20. At 16 weeks the people taking infliximab had a statistically significant improvement in ACR20 compared with those taking placebo (34/52 [65%] of the treatment group achieved ACR20 compared with 5/52 [10%] of the placebo group [RR 6.8; 95% CI, 2.9 to 16.0]). In addition, 24/52 (46%) of the group that had received infliximab achieved an ACR50 response compared with 0% of the placebo group (RR 49.0; 95% CI, 3.1 to 785.1), and 15/52 (29%) achieved an ACR70 response compared with 0% in the placebo group (RR 31.0; 95% CI, 1.9 to 504.9). In terms of PsARC response, 39/52 (75%) patients in the treatment group achieved a response compared with 11/52 (21%) in the placebo group (RR 3.55; 95% CI, 2.05 to 6.13).

4.1.2.3. The results of radiographic assessments of joint disease in the infliximab trial have been published in abstract form. They suggest that at week 50, the total radiographic score had not worsened in 84% of the participants who had received infliximab for the whole period.

4.1.2.4. In total in the IMPACT RCT, 38 of the 52 people in the infliximab group experienced an adverse event (73%) compared with 33 of the 51 people in the placebo group (65%) (RR 1.1; 95% CI, 0.9 to 1.5). Serious adverse events were experienced by one person in each group (RR 1.0; 95% CI, 0.1 to 15.3). Two people in the infliximab group withdrew because of adverse events compared with one person in the placebo group (RR 2.0; 95% CI, 0.2 to 21.4).

4.1.2.5. The results of IMPACT 2 have recently been published. Two hundred patients with active PsA unresponsive to previous treatment were randomised to infusions of infliximab 5 mg/kg or placebo at weeks 0, 2, 6, 14 and 22. All participants had to have shown an inadequate response to current or previous DMARDs or NSAIDs. Concomitant methotrexate and oral corticosteroid use was permitted in the trial.

4.1.2.6. The primary outcome measure in IMPACT 2 was ACR20 at 14 weeks. At 24 weeks the people taking infliximab had a statistically significant improvement in ACR20 compared with those taking placebo (54/100 [54%] of the treatment group achieved ACR20 compared with 16/100 [16%] of the placebo group [RR 3.37; 95% CI, 2.08 to 5.48]). In addition, 41% (41/100) of the group that had received infliximab achieved an ACR50 response compared with 4% of the placebo group (RR 10.25; 95% CI, 3.81 to 27.55), and 27% (27/100) achieved an ACR70 response compared with 2% in the placebo group (RR 13.5; 95% CI, 3.29 to 55.26).

4.1.2.7. In terms of PsARC response, 70/100 (70%) patients in the treatment group achieved a response, compared with 32/100 (32%) in the placebo group (RR 2.19; 95% CI, 1.59 to 2.99). In the subgroups of participants who also had at least 3% of their body surface area affected by psoriasis, the effect of infliximab was statistically significantly greater than that of placebo as measured by the PASI 50 (RR 9.3; 95% CI, 4.5 to 19.1), PASI 75 (RR 52.4; 95% CI, 7.4 to 370.1) and PASI 90 (RR 68.1; 95% CI, 4.2 to 1094.7).

4.1.2.8. In total, by week 24, 100 of the 150 people in a combined group of patients who received infliximab (including those who were randomised to placebo but entered an early escape at week 16 and received infliximab, and those in the placebo group who incorrectly received infliximab) experienced an adverse event (67%) compared with 65 of the 97 people in the placebo group (67%) (RR 0.9; 95% CI, 0.8 to 1.2). Serious adverse events were experienced by 6 people in the placebo group and 13 in the combined infliximab group (RR 1.4; 95% CI, 0.6 to 3.6). Six people in the combined infliximab group withdrew because of adverse events compared with one in the placebo group (RR 3.9; 95% CI, 0.5 to 31.7).

4.2. Cost effectiveness

4.2.1. Published economic evaluations

4.2.1.1. The Assessment Group did not identify any published economic evaluations of either etanercept or infliximab in PsA.

4.2.2. Etanercept - manufacturer’s model

4.2.2.1. The manufacturer of etanercept developed an individual patient-level simulation in their modelling approach using data from one of the RCTs. HAQ scores were used to derive utility gains and estimate the cost of treatments. It was assumed that DMARD treatment with methotrexate and sulfasalazine would have failed before etanercept was considered. People with PsA that failed to respond to etanercept would then be treated with ciclosporin in combination with methotrexate, or leflunomide alone, followed by supportive care on failure. The comparator was the same treatment sequence without etanercept. Costs and benefits were both discounted at 3.5%. The impact of adverse events on costs and utility were not considered in the modelling.

4.2.2.2. Results were presented for four alternative time horizons: 6 months, 2 years, 5 years and 10 years. In the base case, the cost per quality-adjusted life year (QALY) gained for etanercept declined as the time horizon increased, ranging from around £66,590 for a 6-month time horizon to £28,190 for a 10-year time horizon. A Monte Carlo simulation under base-case assumptions and a time horizon of 10 years indicated that 58% of the cost-effectiveness estimates fell below a willingness-to-pay threshold of £30,000, whereas 5% were below £20,000. Univariate sensitivity analyses generated incremental cost-effectiveness ratios (ICERs) ranging from around £35,220 per QALY to £17,200 with a 10-year time horizon.

4.2.3. Infliximab – manufacturer’s model

4.2.3.1. The manufacturer of infliximab developed two Markov models, with the comparator being ‘standard supportive therapy’ only. One model – the Active Joint Model – considered only the short-term effect of flare-ups of active joints, whereas the second – the Chronic Active Joint Model – included this short-term effect but also modelled how flare-ups contribute to the long-term development of deformed joints. The key effectiveness parameters in the models were taken from the IMPACT trial and from an observational study (the Toronto Psoriatic Arthritis Research Program), which also supplied the estimates for utility (based on EQ-5D) and resource use. Disease progression, non-drug costs and people’s responses to treatment were based on the number of active joints. The cost analysis within the model (except the drug costs) was based on resource use estimates from Canada rather than from the NHS. Costs and benefits were both discounted at 3.5%.

4.2.3.2. The models were analysed as ‘first-order’, patient-level simulations comprising 1000 iterations. The analyses used a cost of infliximab of £451.20 per 100-mg vial. The base-case results of the Active Joint Model gave an ICER of infliximab over supportive care of approximately £37,000 (5-year time horizon). The discount rate and time horizon were varied in the sensitivity analyses. Different discount rates appeared to have minimal impact on the ICER of infliximab over supportive care. Two-year, 10-year and 30-year time horizons gave ICERs of around £58,600 (95% CI, £37,660 to £156,460), £33,300 (95% CI, £28,760 to £38,900) and £31,100 (95% CI, £29,640 to £35,840), respectively.

4.2.3.3. In the case of the Chronic Active Joint Model, the ICER in the base-case scenario was around £33,900 (30-year time horizon). Five-year, 10-year and 45-year time horizons gave ICERs of around £41,100 (95% CI, £29,990 to £56,390), £37,400 (95% CI, £28,320 to £46,750) and £35,300 (95% CI, £24,780 to £53,160), respectively.

4.2.4. The Assessment Group model

4.2.4.1. The Assessment Group developed a cohort model, which examined the cost effectiveness of etanercept, infliximab and supportive care (that is, therapy excluding DMARDs and anti-TNFs). It was assumed that these would be the relevant treatment options for people with PsA that had not responded to any DMARD available. Adverse events were not considered in the modelling.

4.2.4.2. The Assessment Group model used short-term trial data (based on a Bayesian evidence synthesis) to model the response of people to anti-TNF therapy at 12 weeks based on PsARC. For those whose condition responded, there was an ongoing risk that they would withdraw from treatment at any time-point. An estimate of annual withdrawal rate was therefore also incorporated into the model. The long-term withdrawal rate was based on a 2-year non-randomised observational study in RA, and assumed a constant rate of withdrawal and no difference between the two biological therapies. People for whom treatment failed at any time-point were assumed to move on to supportive care.

4.2.4.3. As in the manufacturer’s model for etanercept, disability from PsA, as measured by HAQ score, was used as a basis for determining both health-related quality of life (in terms of utility) and the costs of treatment (excluding the unit costs of the biological therapies). People who responded to biological therapy experienced an initial decrease in HAQ score, and a slower progression rate with no further increase in HAQ. The HAQ score of people who failed to respond to biological therapy after the initial 12-week period rebounded, with one of two alternative scenarios. In the best-case scenario after an initial response to therapy failed, the person’s HAQ score rebounded equal to the gain and then increased at the same rate as it would with natural progression. In the worst-case scenario when therapy failed, the person’s HAQ returned to the level and subsequent trajectory at which it would have been had he or she not initially responded to therapy (that is, the increase in HAQ score on failure of therapy exceeded the initial drop).

4.2.4.4. In the analysis, four alternative time horizons were considered: 1 year, 5 years, 10 years and 40 years (lifetime), and uncertainty was explored by probabilistic sensitivity analysis. Under the best-case rebound scenario, the incremental cost, per QALY gained, of etanercept compared with palliative care ranged from £14,820 (women, 40-year time horizon) to £49,370 (men, 1-year time horizon). Under the second, worst-case rebound scenario, the ICERs were higher, ranging from £25,440 (women, 40-year time horizon) to £49,440 (men, 1-year time horizon). The probabilistic sensitivity analysis showed that etanercept and palliative care had the highest probabilities of being cost effective at a threshold of £30,000 per QALY gained.

4.2.4.5. The Assessment Group’s base case model showed that infliximab was consistently dominated by etanercept because of its higher acquisition and administration costs (infliximab did not demonstrate superior effectiveness). This was the case for both rebound scenarios. When etanercept was removed from the comparison, the ICER of infliximab versus supportive care ranged from £21,380 (women, rebound equal to gain, 40-year time horizon) to £90,790 (men, rebound to natural history, 1-year time horizon).

4.2.4.6. A scenario analysis was also undertaken to examine the impact of allowing the HAQ score of responders to etanercept and infliximab to increase at the same rate as natural progression after the initial HAQ decrease. Under this assumption the ICER of etanercept for men over a 10-year time horizon was around £44,600 (base case = £30,400 under the ‘worst-case’ rebound scenario).

4.2.4.7. In addition, the Assessment Group undertook a number of sensitivity analyses relating to the base-case estimates of the effectiveness of infliximab and etanercept (based on PsARC response rates at 12 weeks), the average number of vials of infliximab used in any one infusion, and the non-drug cost of administering infliximab. With the exception of the analysis relating to PsARC response rates, infliximab remained dominated by etanercept in these analyses, even when the non-drug administration costs of infliximab were reduced by more than half.

4.2.4.8. The Assessment Group explored the possibility that there is a better treatment response rate with infliximab in terms of PsARC than was originally estimated in the evidence synthesis. This was done by introducing an alternative prior specification for the study effects: the PsARC response rate at 12 weeks for infliximab was increased to 84% (from 77%) and the PsARC response rate to etanercept was reduced to 73% (from 77%). In terms of cost effectiveness, infliximab was no longer dominated by etanercept, but the ICERs were in excess of £84,000 with a 40-year time horizon (males) and greater than £165,000 with a 10-year time horizon (males).

4.3. Consideration of the evidence

4.3.1. The Appraisal Committee reviewed the evidence available on the clinical and cost effectiveness of etanercept and infliximab, having considered evidence on the nature of the condition and the value placed by users on the benefits of etanercept and infliximab from people with PsA, those who represent them, and clinical experts. It was also mindful of the need to ensure that its advice took account of the effective use of NHS resources.

4.3.2. The Committee noted that the effects of etanercept and infliximab on the inhibition of TNF and the adverse events in studies were similar. Additionally, they accepted that the RCTs demonstrated the efficacy of etanercept and infliximab in people with PsA. The Committee noted, however, that all relevant RCTs were of comparatively short duration and that the sample sizes were generally small. It also noted that the inclusion criteria did not perfectly reflect the population for which these technologies are currently licensed. In particular, in the etanercept RCTs, previous DMARD use was not a requirement for enrolment and IMPACT 2 patients had to have had an inadequate response to either DMARDs or NSAIDs. Expert testimony and analysis of the baseline characteristics of the people included in all the trials, however, suggested that the participants nevertheless represented a population with relatively severe PsA similar to those currently being treated in clinical practice after DMARD failure.

4.3.3. The Committee carefully considered the economic modelling undertaken by the manufacturers and the Assessment Group. It discussed the differences between the modelling strategies and acknowledged their respective merits. It noted the varying time horizons employed in the economic analyses and concluded that although the nature of the condition indicated a potential for long-term benefit from treatment, the short-term trial data currently available meant that it could not confidently support a modelling perspective of greater than 10 years. In addition, the Committee noted the experts’ concerns regarding the limited clinical experience with these agents in the management of PsA and the balance between benefits and risks in the longer term.

4.3.4. The Committee noted that none of the economic models took into account the likely benefits of etanercept and infliximab on the psoriasis component of the disease, and therefore they may have underestimated the relative cost effectiveness of both interventions. The Committee was aware of the limitations of using HAQ scores as a basis for determining health-related quality of life in patients with PsA treated with these agents. Nevertheless, having considered the sensitivity analyses undertaken, which included more than halving the non-drug administration costs of infliximab and increasing the effectiveness of infliximab (see Section 4.2.4.8), the Committee concluded that treatment with etanercept would be more cost effective than treatment with infliximab.

4.3.5. The Committee, however, recognised the benefits of both of these agents and noted that certain subgroups of patients who would not be suitable for treatment with etanercept might benefit from treatment with infliximab. The Committee was persuaded that infliximab should be a treatment option in the circumstances where the use of an anti-TNF agent is considered appropriate in a person with PsA (under the criteria outlined in section 1.1), but where the individual has been shown to be intolerant of, or have contraindications to, treatment with etanercept.

4.3.6. The Committee did not think it appropriate to recommend that a person who did not respond adequately to etanercept (as defined in Section 1.2) should be treated with infliximab. This was due to the absence of data to demonstrate that a person with PsA failing etanercept treatment would benefit from treatment with infliximab. The Committee understood that the British Society for Rheumatology’s Biologics Register was collecting data on sequential use of the anti-TNFs, principally in patients with rheumatoid arthritis. It concluded that such data where relevant should be considered when the current guidance is reviewed.

4.3.7. The Committee further considered issues of concordance with medication related to the different modes of administration of etanercept and infliximab. It was not persuaded by the evidence presented that concordance with treatment would be significantly affected by the different modes of administration of these two agents to a degree that would outweigh the evidence of the cost-effectiveness analysis. The Committee did however appreciate the argument that major difficulties with self-administered injections, for example because of coexisting severe psoriasis, could make this mode of use inappropriate, and infliximab should therefore be considered a treatment option (under the criteria outlined in section 1.1).

4.3.8. The Committee was mindful that the licensed indications for both etanercept and infliximab state that patients must have active and progressive PsA and that there must have been an inadequate response to at least one previous DMARD. However, it accepted the definition of active joint disease and DMARD failure used in the British Society for Rheumatology guidelines as: people must have active joint disease (at least three tender joints and at least three swollen joints) and have failed to respond to adequate therapeutic trials of at least two standard DMARDs.

4.3.9. The Committee considered, on the basis of the trials, that there was no evidence to differentiate between treatment options for mono, oligo or polyarthritis in people with PsA. In addition, it was persuaded in discussion with the clinical experts that evidence for the clinical effectiveness of etanercept and infliximab in people with spinal arthropathy was much less convincing than for peripheral arthritis principally because of under-representation of people with axial disease in the trials.

4.3.10. The Committee discussed the most appropriate method of assessing response to treatment with the anti-TNFs in clinical practice and the time after initiation of therapy that this should be undertaken. The Committee was aware of the limitations of the PsARC measurement criteria, but was persuaded by the clinical experts that PsARC was the most applicable of the possible outcome measures that could be readily used in clinical practice. The Committee further noted the similarity in the numbers of people responding at 12 and 24 weeks in the larger of the etanercept trials, and in the numbers responding at 14 and 24 weeks in the larger of the infliximab trials. The experts advised that 12 weeks was long enough to determine whether a person’s condition was likely to respond to treatment with either etanercept or infliximab. Therefore, the Committee concluded that etanercept should be discontinued if a PsARC response (as defined in Section 2.7) was not achieved at 12 weeks after initiation of therapy. In addition, in the circumstances in which infliximab was to be used, the same criteria should be applied.

4.3.11. The Committee acknowledged that there was the possibility of severe side effects with etanercept and infliximab and that there was little information on the use of these drugs in people with PsA beyond the duration of the RCTs. It was also aware that these drugs could increase the risk of malignancy and that people with PsA were potentially already at an increased risk of skin cancer because of therapeutic exposure to ultraviolet radiation for psoriasis. Consequently, the Committee was of the strong opinion that all people treated with etanercept and infliximab for PsA should be included in a registry to enable ongoing collection of information on long-term outcomes including adverse effects. The Committee considered that cross-referencing (with a psoriasis registry) of the data for people treated for psoriasis and PsA would be necessary to ensure that all information on the use of these agents was captured. The Committee reached the view that treatment with etanercept and infliximab should be initiated and supervised by specialist physicians experienced in the diagnosis and treatment of PsA. Additionally, if a person has both PsA and psoriasis their treatment should be managed by collaboration between a rheumatologist and a dermatologist.