Adalimumab – Manufacturer's (Abbot Laboratories Ltd) Model
The manufacturer's economic evaluation – structured as a patient-based transition-state model – compared the use of adalimumab plus non-steroidal anti-inflammatory drugs (NSAIDs) versus treatment with NSAIDs alone. This model incorporated patient-level data from the Canadian ankylosing spondylitis and ATLAS randomised controlled trials (RCTs), and aimed to simulate treatment decisions based on the British Society for Rheumatology (BSR) guidelines. The trial populations included patients who would not have met BSR eligibility criteria; for example, patients who were intolerant of, or whose ankylosing spondylitis had responded inadequately to, fewer than two NSAIDs.
The model consisted of two components. The first used short-term trial data (first 48 weeks). The second component simulated long-term outcomes for responders for up to 30 years.
In the base-case, the incremental cost-effectiveness ratio (ICER) over a 30-year time horizon was about 23,000 pounds sterling per quality adjusted life year (QALY) gained. Univariate sensitivity analyses on a number of parameters including annual discontinuation rates were undertaken; ICERs varied from 18,000 pounds sterling per QALY gained to around 27,000 pounds sterling per QALY gained (over 30 years).
Etanercept – Manufacturer's (Wyeth Pharmaceuticals) Model
The manufacturer's model compared the use of etanercept plus NSAIDs with NSAIDs alone. The model generated a hypothetical patient population based on patient-level data from two RCTs and an open-label extension. The principal RCT evidence used in the model was drawn from a single study (n = 356). The time horizon was up to 25 years.
In the base-case, the ICER was reported to be around 13,200 pounds sterling over a 25-year time horizon. A number of univariate sensitivity analyses were undertaken. When a utility model based on short-form (SF)-36 data was used, ICERs were found to vary between 17,000 pounds sterling and 70,000 pounds sterling per QALY gained. Probabilistic sensitivity analysis indicated that over a 25-year time period, etanercept has an 88% probability of being cost effective at a threshold willingness to pay of 15,000 pounds sterling.
Infliximab – Manufacturer's (Schering-Plough Ltd) Model
The manufacturer's model is based on a combined decision tree and Markov chain structure, and compares infliximab versus 'standard therapy'. Two analyses were described, one based on the 24-week outcomes of the ASSERT trial and the other on a smaller study of up to 12 weeks. The placebo groups in these studies were assumed to have received standard therapy as these studies allowed the concomitant use of NSAIDs.
In the base-case, the reported ICERs in the original submission were under 20,000 pounds sterling. However, this was based on an inaccurate model, which in part allowed patients who withdrew from infliximab treatment to avoid being assigned an 'off-treatment' disease progression. On correcting this error, the manufacturer reported base-case 70-year ICERs of approximately 27,000 pounds sterling to 28,000 pounds sterling per QALY gained (depending on which of the two studies is used to inform the calculation). In contrast, the Assessment Group found that, on correcting the model within an Excel replica, the lifetime ICERs were between 41,000 pounds sterling and 50,000 pounds sterling per QALY gained. The manufacturer also reported corrected ICERs for the scenario in which disease progression while on treatment is assumed to be 50% of natural history (that is, 0.035 units per year), and the ICERs rise to between 34,000 pounds sterling and 35,000 pounds sterling per QALY gained.
The Assessment Group Model
The Assessment Group examined the use of adalimumab, etanercept and infliximab compared with 'conventional treatment'. 'Conventional treatment' was defined in terms of the placebo arms of two adalimumab RCTs. The group explored the cost effectiveness of these interventions over the short term (1 year) and over a time horizon of up to 20 years.
Under base-case assumptions, from week 30 onwards it was assumed that spontaneous recovery without treatment would occur at a rate of 17.1% as identified in the patient-level analysis of two adalimumab RCTs supplied in the Abbott submission. This assumption was explored in univariate and multivariate sensitivity analyses. In univariate sensitivity analyses, in which it was assumed there was no spontaneous recovery in the placebo arm, the ICERs for adalimumab and etanercept over a 20-year time horizon decreased from 92,000 pounds sterling (base-case) to 57,000 pounds sterling. The ICER for infliximab decreased from 168,000 pounds sterling (base-case) to 109,000 pounds sterling.
Univariate and multivariate sensitivity analyses were undertaken. Multivariate sensitivity analyses identified scenarios in which adalimumab/etanercept could be considered cost effective, with ICERs ranging from 12,000 pounds sterling to 118,000 pounds sterling. Important factors influencing the long-term cost effectiveness of these two drugs included assumptions about spontaneous recovery, withdrawal rate from treatment and the Bath Ankylosing Spondylitis Functional Index (BASFI) progression rate. Multivariate sensitivity analyses on the infliximab results identified no scenario in which the ICER dropped below 35,000 pounds sterling.
Further Analysis by the Decision Support Unit (DSU)
Following consultation on the submissions from the three manufacturers and the Assessment Group, the Committee requested additional analysis to be carried out by the Decision Support Unit to identify reasons for the large differences in the cost-effectiveness results and to determine whether the differences in the results still existed when an agreed set of common parameter values were included.
Using a common set of parameter values in the manufacturers' models and applying the assumption of no disease progression after 1 year for tumour necrosis factor (TNF)-alpha inhibitor treatment responders to the Assessment Group's model, gave revised results for adalimumab/etanercept of 30,000 pounds sterling per QALY gained, down from 42,000 pounds sterling per QALY gained. If no disease progression was assumed for adalimumab or etanercept after 20 weeks, the ICER becomes 22,000 pounds sterling per QALY gained. The equivalent ICER for infliximab if no disease progression was assumed after 20 weeks was 49,000 pounds sterling per QALY gained. The DSU commented that the assumption of zero response in the placebo arm was a favourable one. If this assumption is not made, the ICERs for the Assessment Group's model move from 30,000 pounds sterling per QALY gained to 47,000 pounds sterling per QALY gained if no disease progression is assumed after the first year, and from 22,000 pounds sterling per QALY gained to 31,000 pounds sterling per QALY gained, if no disease progression is assumed after the first 20 weeks.
Consideration of the Evidence
The Committee considered the evidence for the cost effectiveness of TNF-alpha inhibitors.
The Committee considered the DSU's assumption of stable Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) and BASFI after 20 weeks to be appropriate in this context. The resulting ICER of 22,000 pounds sterling per QALY gained calculated by the DSU for adalimumab and etanercept using the Assessment Group model, in which BASDAI and BASFI remain stable after 20 weeks, was also considered to be plausible in this context. The Committee noted that the ICER for infliximab, using the stable BASDAI and BASFI profiles, was 49,000 pounds sterling per QALY gained.
The Committee considered that the assumption around no spontaneous resolution of symptoms, equivalent to absence of a placebo response, was unlikely. It heard from the DSU that if a 17% response in the placebo arm was assumed (as in the original Assessment Group model) then the ICER of 22,000 pounds sterling per QALY gained would increase to 31,000 pounds sterling per QALY gained for etanercept and adalimumab. The Committee considered that these two figures represented a reasonable range of cost effectiveness based on the evidence. The equivalent figures for infliximab were 49,000 pounds sterling to 65,000 pounds sterling per QALY gained. Therefore, on balance, taking into account all of its previous assumptions, the Committee concluded that adalimumab and etanercept for the treatment of severe ankylosing spondylitis could be considered a cost-effective use of National health Service (NHS) resources in the context of achieving a continued response to treatment.
The Committee discussed the cost effectiveness of infliximab in further detail. Because the available evidence persuaded the Committee that infliximab was not cost effective in treating ankylosing spondylitis, it concluded that it could not recommend the use of infliximab simply on the basis of another treatment choice.
Refer to Section 4 of the original guideline document for details of the economic analyses provided by the manufacturers, the Assessment Group comments, and the Appraisal Committee considerations.