Etanercept, infliximab and adalimumab for the treatment of psoriatic arthritis

The 2 double-blind, placebo-controlled RCTs of etanercept in adults with active psoriatic arthritis were Mease 2000 (n=60; follow-up 12 weeks) and Mease 2004 (n=205; follow-up 24 weeks). In both trials, 25 mg etanercept was administered by subcutaneous injection twice a week. The inclusion criteria for both trials were active psoriatic arthritis (defined as more than 3 swollen joints and more than 3 tender or painful joints, although only the more recent trial specified stable plaque psoriasis), and psoriatic arthritis that had not responded adequately to non-steroidal anti-inflammatory drugs (NSAIDs). The primary outcome variable in the Mease 2000 trial was Psoriatic Arthritis Response Criteria (PsARC) and in Mease 2004 it was ACR 20. Data for Psoriasis Area and Severity Index [PASI] at week 12 were available from Mease 2000 only.

The Assessment Group conducted a meta-analysis of the outcomes for etanercept at 12 weeks and the pooled estimates from both trials showed that etanercept was statistically significantly more effective than placebo for all outcomes (PsARC, ACR 20, ACR 50, ACR 70, and Health Assessment Questionnaire [HAQ] percentage change from baseline). For PsARC, the pooled relative risk (RR) estimate was 2.60 (95% confidence interval [CI] 1.96 to 3.45), with some evidence of statistical heterogeneity (I²=34%) between the 2 studies' estimates. For PASI 50, the results from the Mease 2000 trial at 12 weeks showed that etanercept was more effective than placebo (RR=2.00 [95% CI 0.72 to 5.53]) although this was not statistically significant. For PASI 75 the results showed that etanercept was statistically significantly more effective than placebo (RR=11.00 [95% CI 0.65 to 186.02]; p=0.0154).

At 24 weeks the treatment effect for all joint disease outcome measures was statistically significantly greater for etanercept than for placebo, though these data were available only for 1 trial, Mease 2004. At 24 weeks, the annualised rate of progression as measured radiologically using the Total Sharp Score was statistically significantly lower in people treated with etanercept than in people treated with placebo (Total Sharp Score -0.56; 95% CI -0.86 to -0.26).

At 24 weeks the treatment effect on psoriasis favoured etanercept with RRs for PASI 75 of 7.05 (95% CI 1.68 to 29.56), PASI 50 of 2.65 (95% CI 1.46 to 4.80) and PASI 90 of 1.88 (95% CI 0.36 to 9.90). At 1 year the mean annualised rate of progression on the Total Sharp Score for all people was -0.03 (standard deviation [SD] 0.87), indicating that on average there was no clinically significant progression of joint erosion based on uncontrolled follow-up data.

The 2 double-blind, placebo-controlled RCTs of infliximab for the treatment of psoriatic arthritis were IMPACT and IMPACT 2. In the IMPACT trial, participants (n=104) were randomised to receive infusions of infliximab (5 mg/kg) or placebo at weeks 0, 2, 6 and 14 with follow-up at week 16. In IMPACT 2, people (n=200) were randomised to receive infusions of placebo or infliximab 5 mg/kg at weeks 0, 2, 6, 14 and 22, with assessments at weeks 14 and 24. In both RCTs the inclusion criteria required that participants' psoriatic arthritis should have 5 or more swollen/tender joints, and that their disease had an inadequate response to at least 1 disease-modifying antirheumatic drug (DMARD). IMPACT 2 also required people to have active plaque psoriasis with at least 1 qualifying target lesion (2 cm or more in diameter).

The Assessment Group conducted a meta-analysis of the outcomes for infliximab at 14 weeks and the results for both trials reported a statistically significant improvement in PsARC for people receiving infliximab, relative to those receiving placebo (pooled RR 3.44, 95% CI 2.53 to 4.69). There was some evidence of statistical heterogeneity (I²=68%) between the 2 study estimates. Infliximab was statistically significantly more effective than placebo for all pooled estimates for outcomes of joint response (ACR 20, ACR 50 and ACR 70) as well as the pooled percentage change from baseline in HAQ score with infliximab compared with placebo (mean difference -60.37 [95% CI -75.28 to -45.46]).

The Assessment Group also presented pooled estimates for the outcomes of the skin component of psoriatic arthritis over 14 to 16 weeks and the results showed that infliximab was statistically significantly more effective than placebo.

The IMPACT 2 trial was randomised for 24 weeks followed by an open-label period. The data for all measures of joint disease, psoriasis and HAQ were similar to those at the 14-week follow-up, suggesting that infliximab's benefits were maintained for up to 24 weeks of treatment and for longer-term follow-up (50 weeks for IMPACT and 54 weeks for IMPACT 2) although the data for the longer-term follow-up were uncontrolled.

In terms of radiographic assessment, there was no statistically significant change from baseline in the total modified van der Heijde-Sharp Score for infliximab-treated people followed up at 50 weeks in the IMPACT trial (n=70; -1.72 [5.82]) or 54 weeks in the IMPACT 2 trial (infliximab or infliximab -0.94 [3.4]; placebo or infliximab 0.53 [2.6]), suggesting infliximab may inhibit progression of joint damage. However, as with other outcomes measured after week 24, there was no placebo group for comparison.

The 2 double-blind, placebo-controlled RCTs of adalimumab in adults with active psoriatic arthritis were ADEPT (n=313, follow-up of 24 weeks) and Genovese 2007 (n=100, follow-up of 12 weeks). In both trials adults were randomised to adalimumab (40 mg every other week) or placebo. The inclusion criteria for both RCTs required people to have active psoriatic arthritis (defined in both trials as more than 3 swollen joints and more than 3 tender or painful joints, with active psoriatic skin lesions or a documented history of psoriasis). Overall, the baseline characteristics demonstrated that the trial populations were indistinguishable and represented people who required DMARDs or therapy with tumour necrosis factor (TNF) inhibitors.

The Assessment Group conducted a meta-analysis of the outcomes for adalimumab at 12 weeks and the results from both trials showed a statistically significant improvement for adalimumab compared with placebo for all outcome measures. The pooled RR for PsARC was 2.24 (95% CI 1.74 to 2.88) and the pooled RR for ACR 20 was 3.65 (95% CI 2.57 to 5.17). The pooled RRs for ACR 50 and ACR 70 also favoured adalimumab, although their related CIs were wide. Regarding the associated skin disease, 12‑week PASI response measures were reported by only 1 trial (ADEPT), and the response was statistically significantly greater for adalimumab than placebo at all 3 PASI thresholds: PASI 50 RR=5.00 (95% CI 2.77 to 9.03); PASI 75 RR=11.33 (95% CI 3.65 to 35.17); and PASI 90 RR=43.00 (95% CI 2.66 to 696.04). The CIs, especially for PASI 75 and PASI 90, were wide.

The ADEPT trial was randomised for 24 weeks. The data for all measures of joint disease, psoriasis and HAQ were similar to those at 12-week follow-up. In addition, this trial also reported a statistically significant difference in mean change in Total Sharp Score from baseline (-0.2 versus 0.1, p<0.001) favouring adalimumab over placebo in terms of delayed progression of joint disease, although this duration of follow-up is short.

In the absence of head-to-head RCTs on the relative efficacy of the 3 TNF inhibitors, an indirect comparison was undertaken by the Assessment Group using placebo as the common comparator. The results were expressed as the probability of each of the TNF inhibitors achieving a response for the outcome measures PsARC, HAQ, PASI and ACR. Infliximab was associated with the highest probability of achieving a response for all of the outcomes measured. The probability of response in joint disease (PsARC and ACR) was higher with etanercept than with adalimumab, and the probability of response in skin disease (PASI) was higher with adalimumab than with etanercept.

There were no RCTs that directly compared the 3 drugs. To evaluate the adverse events of the 3 TNF inhibitors the Assessment Group reviewed a range of study types including RCTs, open-label extensions of trials and observational studies.

The Assessment Group provided a range of estimates for serious adverse event and withdrawal rates across non-randomised studies and large RCTs. These comprised serious infections, cancer, activation of latent tuberculosis, mortality and withdrawals from treatment because of adverse events.

The Assessment Group acknowledged that the adverse event data were primarily from people with rheumatoid arthritis or other indications, so it is unclear to what extent these can be generalised to psoriatic arthritis. Overall, the limited evidence prevented them from drawing firm conclusions from the systematic review about the comparative adverse event profile of the 3 TNF inhibitors.

The Assessment Group performed a systematic review of published literature and identified 3 studies (Bansback et al. 2007; Bravo Vergel 2006; and Olivieri et al. 2008) that met the inclusion criteria for the cost-effectiveness review.

The study by Olivieri et al. (2008) was difficult to compare with the other studies because in this study all TNF inhibitors were considered as a group compared with DMARDs. There were no model results. The economic evaluation was made using before-and-after studies and the effectiveness evidence was based on a single trial. This produced an incremental cost-effectiveness ratio (ICER) of around €40,000 (£34,700) per quality-adjusted life year (QALY) gained for TNF inhibitors.

The study by Bansback et al. (2007) compared etanercept with ciclosporin and leflunomide. The economic model focused on response according to PsARC and associated Health Assessment Questionnaire (HAQ) score, with changes in HAQ and further withdrawals modelled over 10 years. Mease 2004 was the source of evidence for response rates and HAQ. The base-case results showed an ICER of around £28,000 per QALY gained for etanercept compared with ciclosporin and £38,000 per QALY gained for etanercept compared with leflunomide.

The study by Bravo Vergel (2006) compared etanercept with infliximab and palliative care. The model included response according to PsARC and associated HAQ score. Changes in HAQ and further withdrawals were modelled over 40 and 10 years. Evidence from Mease 2000, Mease 2004 and IMPACT was used to model the PsARC response. The ICER for etanercept was between £26,361 and £30,628 per QALY gained compared with palliative care depending on the assumptions made about the deterioration in HAQ score at treatment withdrawal (rebound). Infliximab was the most effective strategy, and generated the highest number of QALYs.

A published cost-effectiveness model originally used to support a submission to NICE in 2004 was adapted to incorporate additional effectiveness evidence and new comparators. The adjusted model compared the costs and benefits associated with etanercept, infliximab, adalimumab and best supportive care over a lifetime horizon. Best supportive care was assumed to be ciclosporin because the population considered in the model were assumed to have already tried other DMARDs (leflunomide, sulfasalazine and methotrexate).

The base-case results showed that the costs for best supportive care were £53,860 with QALYs of 5.96, and for etanercept the costs were £65,650 with QALYs of 6.90. This resulted in an ICER of £12,480 per QALY gained for etanercept when compared with best supportive care. Adalimumab was extendedly dominated by a combination of etanercept and palliative care (that is, additional QALYs could be generated with etanercept relative to adalimumab at a lower cost per QALY gained than is generated by adalimumab relative to palliative care). Infliximab was dominated by adalimumab (that is, infliximab was more costly and less effective than adalimumab).

In the economic analysis submitted by the manufacturer of infliximab 4 treatment alternatives were compared over a lifetime horizon. These included maintenance treatment with a TNF inhibitor (infliximab, adalimumab or etanercept) followed by a sequence of DMARDs. The comparator was palliative care with DMARDs. For the health-economic model, the incremental treatment effects for the comparative treatments were estimated for infliximab, etanercept and adalimumab. The direct drug costs for the TNF inhibitors were obtained from BNF edition 56.

The manufacturer presented base-case results for 3 different scenarios: people weighing 60 kg, 70 kg with vial optimisation for infliximab treatment (that is, making local arrangements so that vials can be shared between patients who are being treated with infliximab, reducing wastage) and 80 kg. For people weighing 60 kg the base-case results showed that infliximab produced an ICER of £16,942 per QALY gained when compared with palliative care. For people weighing 70 kg, and accounting for vial optimisation, infliximab produced an ICER of £19,982 per QALY gained versus palliative care. For people weighing 80 kg infliximab produced an ICER of £23,022 per QALY gained when compared with palliative care.

The manufacturer of adalimumab used an individual sampling model to simulate the disease progression of a cohort of people with psoriatic arthritis over a lifetime horizon under different treatment sequences. A 3-month cycle was used. Baseline characteristics from the ADEPT trial for people for whom 2 previous DMARDs had failed were used in the base-case analysis. The cost of all drugs used in the analysis was calculated based on the recommended dosages and vial prices given in the Monthly Index of Medical Specialties. The model assumed that 4 × 100‑mg vials of infliximab were required per infusion, based on an average person weighing 80 kg.

The base-case results showed that adalimumab, with a mean cost of £73,072 and QALYs of 8.33, was the most cost-effective treatment strategy when compared with a DMARD (mean costs of £47,537 and QALYs of 7.47), resulting in an ICER of £29,827 per QALY gained. Etanercept was more costly and had the same mean QALYs gained as adalimumab (8.33). Infliximab was more costly and more effective than adalimumab, which resulted in an ICER of £199,596 per QALY gained compared with adalimumab.

The Assessment Group updated the economic model developed for 'Etanercept and infliximab for the treatment of adults with psoriatic arthritis' (NICE technology appraisal 104). This model allowed the 3 TNF inhibitors to be compared with each other. A probabilistic decision analytic model was developed to estimate the incremental costs and incremental QALYs of the 3 TNF inhibitors compared with palliative care over a lifetime horizon (40 years), only. The price year was 2008/2009 and costs and benefits were discounted at a rate of 3.5%.

The decision analytical model followed a cohort of people that represented the average characteristics of participants in the RCTs and had a Markov structure. People in the cohort were assumed to be 47 years old, had been diagnosed with psoriatic arthritis 7 years previously, were assumed to weigh 60 kg to 80 kg, and had psoriatic arthritis that had inadequately responded to at least 2 DMARDS. People in the treatment arm received etanercept, infliximab or adalimumab and people in the control arm received palliative care. The disease's response to treatment was assessed between 12 and 16 weeks. It was assumed that people whose disease had responded to treatment stayed in the treatment arm, while treatment was discontinued in people whose psoriatic arthritis failed to adequately respond to treatment – these people were assumed to go on to receive palliative care.

The following assumptions were included in the Assessment Group's model: people in the initial 3-month trial period had some improvement in HAQ (even if they did not reach the PsARC threshold); people who had a PASI 75 response would gain at least a 75% improvement in psoriasis compared with baseline PASI; people continuing on TNF inhibitors maintained their initial improvement in HAQ; and the same ongoing risk of withdrawal from treatment was used for all TNF inhibitors (withdrawal because of reduction in efficacy, adverse events or other reasons).

The base-case analysis in the Assessment Group's model assumed a lifetime (40-year) time horizon for costs and QALYs, a baseline HAQ of 1.05, a baseline PASI of 7.5, rebound equal to gain, and incorporate the correlation between PsARC and PASI 75 outcomes. Health utility was measured as a function of HAQ and PASI based on linear regressions of EQ5D utility versus HAQ and PASI provided by the manufacturers based on RCT evidence. The total lifetime discounted health associated with palliative care was about 5.2 QALYs because the base case assumed that utility declined fairly rapidly in people with uncontrolled arthritis, and may have been less than 0 (representing a health state worse than death) in later years.

The base-case model assumed that people's psoriatic arthritis had failed to respond to treatment with at least 2 DMARDS but they had not received previous treatment with TNF inhibitors. The Assessment Group also modelled the cost effectiveness of sequencing TNF inhibitor therapies after people's psoriatic arthritis failed to respond to a first-line TNF inhibitor. The base-case analysis reported the lifetime costs and QALYs of the 3 TNF inhibitors in people with mild-to-moderate psoriatic arthritis, which was presented as an incremental analysis ranking the alternative strategies by mean cost.

Following comments made by NICE consultees on the Technology Assessment Report and model of December 2009, the Assessment Group revised the cost-effectiveness analysis results. The Assessment Group took into account the manufacturer of adalimumab's revised estimates from their RCTs of the effect of adalimumab on HAQ change for PsARC responders and non-responders. The Assessment Group corrected a standard error calculation when extracting data for the evidence synthesis and used the correct calculation of the costs of adalimumab and etanercept. The results for the base case showed that infliximab was the most effective treatment taking into account both joint and skin effects (QALYs of 7.3), followed by etanercept (QALYs of 7.0), then adalimumab (QALYs of 6.6). Infliximab was also the most costly treatment (£88,442), followed by etanercept (£74,841), then adalimumab (£68,638). The ICER of etanercept compared with palliative care was £17,853 per QALY gained. The ICER for infliximab compared with etanercept was £44,326 per QALY gained. Adalimumab was extendedly dominated by a combination of etanercept and palliative care (that is, additional QALYs could be generated with etanercept relative to adalimumab at a lower cost per QALY gained than the ICER of adalimumab relative to palliative care, adalimumab was therefore excluded from the incremental analysis). Etanercept had the highest probability of being cost effective with probabilities of being cost effective of 44% if the maximum acceptable amount to pay for an additional QALY was £20,000 and 48% if the maximum acceptable amount to pay for an additional QALY was £30,000.

The Assessment Group conducted several univariate sensitivity analyses using different sets of assumptions. The Assessment Group presented the results according to whether the ICER was less than £20,000 per QALY gained, between £20,000 and £30,000 per QALY gained or greater than £30,000 per QALY gained.

The results of these analyses suggested that the ICER of etanercept increased to above £20,000 per QALY gained or was dominated by other strategies when the following assumptions were used and all other variables take mean values as in the base case:

• A patient treated for psoriatic arthritis whose skin disease does not achieve a PASI 75 response is admitted to hospital for treatment of psoriasis (annual treatment). The base case assumed these patients are offered ultraviolet (UV) light therapy.

• The HAQ rebounds after withdrawal from TNF inhibitors to natural history rather than to initial gain.

• Treatment with TNF inhibitors becomes ineffective (relative to no treatment) after 10 years.

• Infliximab requires 3 vials rather than 4 vials per administration.

• All responders to PsARC have the same change in HAQ at 3 months, regardless of the TNF inhibitor used.

For most sensitivity analyses performed by the Assessment Group, the ICER for infliximab was greater than £30,000 per QALY gained. The ICER of infliximab fell below £30,000 per QALY gained, when the following assumptions were used and all other variables take mean values as in the base case:

• A patient treated for psoriatic arthritis whose skin disease does not achieve a PASI 75 response is admitted to hospital for treatment of psoriasis (annual treatment). The base case assumed these patients are offered UV light therapy.

• Infliximab requires 3 vials rather than 4 vials per administration.

• If the manufacturer of infliximab's estimates of the cost of treating psoriasis with UV light therapy are used in the Assessment Group's model.

• HAQ improves while on biological therapy. The base case assumes no change after the first 3 months.

The ICER of adalimumab fell below £20,000 per QALY gained and was no longer dominated by other strategies, when the following assumptions were used and all other variables take mean values as in the base case:

• All responders to PsARC have the same change in HAQ at 3 months, regardless of the TNF inhibitor used.

• A patient treated for psoriatic arthritis whose skin disease does not achieve a PASI 75 response is admitted to hospital for treatment of psoriasis (annual treatment). The base case assumed these patients are offered UV light therapy.

• If the manufacturer of infliximab's estimates of the cost of treating psoriasis with UV light therapy are used in the Assessment Group's model.

The Assessment Group performed a sensitivity analysis assuming all TNF inhibitors had the same change in HAQ benefit at 3 months for a PsARC responder. The Assessment Group calculated that the ICERs per QALY gained were £17,717 for adalimumab compared with palliative care, £22,056 for etanercept compared with adalimumab and £50,806 for infliximab compared with etanercept.

The Assessment Group also provided cost-effectiveness results for subgroups with different patient characteristics. For a cohort in which baseline PASI was moderate to severe (PASI of 12.5 instead of 7.5 as in the base-case) the ICER of adalimumab versus palliative care was £16,310 per QALY gained, the ICER of etanercept versus adalimumab was £19,319 per QALY gained and the ICER of infliximab versus etanercept was £27,778 per QALY gained. For a cohort of people with negligible baseline psoriasis etanercept was the most cost-effective strategy with an ICER of £18,512 per QALY gained compared with palliative care, the ICER of infliximab compared with etanercept was £64,744 per QALY gained and adalimumab was extendedly dominated by a combination of etanercept and palliative care. For a cohort of people with moderate-to-severe psoriasis (baseline PASI of 12.5) whose disease did not achieve a PASI 75 response and are assumed to be admitted to hospital for treatment of psoriasis (annual treatment) instead of annual UV light therapy, the ICER for adalimumab compared with palliative care was £7901 per QALY gained, the ICER for infliximab compared with adalimumab was £10,636 per QALY gained and etanercept was dominated by (that is, was more costly and generated less QALYs than) infliximab.

The Assessment Group presented an additional analysis in which people were assumed to continue on biological treatment after 3 months if their disease had either an adequate PsARC or a PASI 75 response (base case: PsARC only). For etanercept compared with palliative care the ICER was £17,859 per QALY gained, the ICER for infliximab compared with etanercept was £38,194 per QALY gained and adalimumab was extendedly dominated by a combination of etanercept and palliative care (that is, additional QALYs could be generated with etanercept relative to adalimumab at a lower cost per QALY gained than the ICER of adalimumab relative to palliative care).

The Assessment Group presented an analysis that compared the sequencing of the different TNF inhibitors in people with mild-to-moderate skin disease if a first TNF inhibitor has failed. The ICERs depended on which drug was used as first-line therapy, and was therefore ineligible for use as second-line therapy. The Assessment Group noted that the ICERs were broadly similar for people whose psoriatic arthritis failed to respond to first-line therapy because of adverse effects and those whose disease failed first-line therapy because of inefficacy.

An additional sensitivity analysis was performed by the Assessment Group at the Committee meeting and subsequently confirmed by running the model probabilistically. This analysis assumed that adalimumab and etanercept were equally effective while the PsARC responses for infliximab remained the same as in the original analysis (that is, infliximab was assumed to be more effective than adalimumab and etanercept). The ICER for both adalimumab and etanercept compared with palliative care was £18,296 per QALY gained and the ICER for infliximab compared with adalimumab and etanercept was £45,557 per QALY gained.