EDITORIAL |
Infliximab, etanercept and adalimumab for the treatment of ankylosing spondylitis: cost-effectiveness evidence and NICE guidance
1Health Economics and Decision Science, ScHARR, University of Sheffield, UK and 2Centre for Health Evaluation and Outcome Sciences, University of British Columbia, Vancouver, Canada
Correspondence to: A. Wailoo, Health Economics and Decision Science, ScHARR, Regent Court, 30 Regent Street, Sheffield S1 4DA, UK. E-mail: a.j.wailoo{at}sheffield.ac.uk
Background
Recent editions of this journal have seen papers published that report on the cost-effectiveness of etanercept [1] and adalimumab [2] in AS. These join similar analyses for infliximab [3, 4]. Each of these studies was commissioned by the drug manufacturers (Wyeth, Abbott and Schering Plough, respectively) and are consistent in terms of their conclusions: each of the TNF-
inhibitors is likely to be cost-effective in the UK National Health Service (NHS) compared with conventional therapy. Readers might therefore expect the National Institute for Health and Clinical Excellence (NICE), which is currently considering the use of these treatments for AS in the NHS and has received cost-effectiveness evidence from all three manufacturers, to issue guidance that recommends the use of these drugs in the NHS. However, the issues that the NICE Technology Appraisals Committee have had to consider are not as straightforward as implied by these three publications. This editorial seeks to highlight these issues.
NICE's consideration of the biologic drugs for AS has lasted a marathon 2 years so far and has considered evidence from all three manufacturers, various other stakeholders and a report by an independent assessment group (AG) based at the University of Liverpool, UK [5]. Five meetings of its Appraisal Committee have been held compared with the more usual one or two meetings for most other technologies. Perhaps uniquely, NICE also commissioned additional analyses of the manufacturer and independent assessment group submissions from its Decision Support Unit (DSU) [6, 7] in order that apparent differences in results could be identified and considered in formulating NHS guidance.
At the root of the need for these additional analyses was the requirement for the Institute to understand the basis for the vast differences in cost-effectiveness estimates generated by the manufacturers and the independent AG. At the outset of DSU involvement, estimates of the incremental cost-effectiveness ratios (ICERs) for etanercept were £22 700/Quality Adjusted Life Year (QALY) (Wyeth) vs £98 910/QALY (AG), for adalimumab were £23 097/QALY (Abbott) vs £98 910/QALY (AG) and for infliximab were £19 196/QALY (Schering Plough) vs £175 000/QALY (AG). To give context to these numbers, an ICER of <£30 000/QALY is usually considered by NICE to be within an acceptable range.
General features of the analyses
As with many chronic diseases, a key challenge in estimating cost-effectiveness lies in the approach to extrapolating beyond short-term clinical trials (typically 12 or 24 weeks in AS) to the time period over which differences in both costs and health benefits can reasonably be expected to occur. As with many of the rheumatic diseases, the appropriate time period to consider in AS is long-term and may be the rest of a patient's life. Mathematical models designed to make these long-term estimates need to reflect all appropriate data and, if no such data exist, must be based on explicit and defensible assumptions.
Broadly, the modelling approaches taken by the three manufacturers were similar. All the models estimate long-term patient profiles in terms of Bath AS Functional Index (BASFI) and Bath AS Disease Activity Index (BASDAI) that are subsequently used to estimate QALYs and costs. Patients receiving treatment are classed as either responders or non-responders according to British Society for Rheumatology (BSR) guidelines based on data from the relevant clinical trials. Non-responders are withdrawn from treatment whilst responders are considered to maintain improvements in BASFI and BASDAI, until they withdraw from treatment. At withdrawal, patients experience some degree of worsening in BASFI and BASDAI.
In our detailed examination of each of the manufacturer models and that submitted by the independent AG, and consideration of the evidence by the NICE Appraisal Committee, the following were identified as important issues even after a number of common parameter values such as withdrawal rates, time horizon and baseline BASDAI/BASFI were implemented.
What happens to BASDAI/BASFI over the long term for patients who remain on therapy?
The base case for each manufacturer model is based on an assumption that there is no progression in either BASFI or BASDAI whilst patients remain on treatment. Whilst this assumption has been generally accepted for biologic treatment in rheumatoid arthritis based on the evidence of retardation for both radiological and functional progression, such evidence of disease modification in AS is lacking [8]. Both the analyses by Ara et al. [1] and Kobelt et al. [4] rely on data from open-label extensions to the trials of etanercept and infliximab, respectively. Notwithstanding concerns over selection bias and the lack of a comparator in this type of study extension, these particular examples are based on relatively short-term data and very small numbers of patients. The analysis by Botteman et al. [2] does not offer any data to support this assumption.
Worsening of either BASDAI and BASFI whilst on treatment could increase the estimates of cost-effectiveness substantially. For example, using progression rates estimated from the trials of adalimumab by the independent assessment group, produce estimates of the ICER in excess of £40 000/QALY for etanercept and adalimumab and in excess of £80 000/QALY for infliximab.
What happens to BASDAI/BASFI when patients withdraw?
All the models assume that patients rebound back to baseline BASDAI after they withdraw from biologic treatment. Since all models also assume no worsening in the untreated comparator group, both sets of patients are therefore equivalent in terms of BASDAI from the time of withdrawal. The assumption regarding the magnitude of BASFI rebound is more problematic. If patients who receive no treatment are worsening over time, then there are different approaches that can be taken when patients withdraw. This is illustrated in Fig. 1 using data from Kobelt et al. [4]. Two possibilities are that either the patient rebounds to a level equivalent to the untreated group, or the patient rebounds to the original baseline level. The latter allocates a benefit to treatment that lasts for the rest of the patient's lifetime, but there is no additional cost because the patient has withdrawn from biologic treatment. It is this second approach that all the analyses follow to some extent.
|
The analysis by Kobelt et al. [4] implements an assumption of rebound to the level of the untreated cohort. However, close investigation of the model revealed a significant error in implementation. While the model assumes a rebound to the original level after withdrawal, no further deterioration in BASFI over the remainder of the patients’ lifetime was implemented to the infliximab group in contrast to the comparator arm. The further benefits allocated to patients in the infliximab group were significant, and after correcting for this we estimated the actual ICER of the Kobelt et al. [4] model to be £59 430/QALY compared with the £19 196/QALY reported in the publication.
In addition, the independent AG identified concerns with the analysis of trial data for adalimumab undertaken by Botteman et al. [2], which informs the BASDAI and BASFI profiles over time in the cost-effectiveness model. Patients in the clinical trials were able to choose an ‘early escape’ open-label treatment with adalimumab. However, those exercising this choice in the placebo and treatment arms are not treated equally, giving the impression of larger treatment benefits than may be warranted.
Overview
Consequently, after much scrutiny by the DSU a range of values of the ICER for each treatment were presented to the NICE Committee based on a set of assumptions that they regarded as reasonable [6].
Readers might be surprised that NICE makes judgements that can have such important implications on the lives of patients with rheumatological conditions, using models that extrapolate what is relatively short-term data to long-term outcomes. Without these extrapolations, the potential benefits of many rheumatological interventions are underestimated, and consequently, when compared with other diseases that have more immediate effects (e.g. cancer survival), might not be considered a worthy use of health care resources.
Of course, these extrapolations are subject to uncertainty that is routinely examined by NICE as part of the appraisal process. In this instance, two issues of importance for the rheumatology community were identified.
First, and as noted by Ara et al. [1], there is a dearth of data in relation to several important factors in AS. Partly, this is a result of the short time in which the treatments have been available, and it is important to note that NICE issues a review date alongside its guidance that is linked to the production of new data. One purpose of the NICE analyses is to highlight the parameters whose uncertainty is most influential on the economic case for therapy. It is of crucial importance that the rheumatological community and the manufacturers consider how to address these current uncertainties for that review by appropriate long-term studies, including long-term progression whilst on treatment.
Second, extensive peer review of cost-effectiveness articles submitted for publication is difficult. Reviewers typically work for no fee and frequently do not have access to the electronic model to enable the implementation of the model to be investigated. Furthermore, whilst several generic checklists for reporting cost-effectiveness studies exist [9, 10], these do not and cannot address disease specific issues in modelling. These considerations have led to the development, by the OMERACT research group, of a reference case for economic evaluations in AS. The authors would like to recommend the use of this reference case to the rheumatology community in assessing further economic evaluations in this disease area [11].
In this case, the important issues of concern were found only via the rigorous NICE appraisal process. This same process examined all four submitted models, including that developed by the independent AG, to allow the Appraisal Committee to understand the workings and implied assumptions of each. Whilst many characterize cost-effectiveness analysis as a ‘black-box’ approach to decision making, we believe that NICE has gone to great length to ensure that its consideration of the TNF- inhibitors for the treatment of AS does not display this characteristic. It is our hope that this explanation will better describe the evidence reviewed by the NICE Committee, and some of the difficulties faced by NICE when appraising new health technologies.
Acknowledgement
Funding: This study was funded by the National Institute for Health and Clinical Excellence, Decision Support Unit.
Disclosure statement: The authors have declared no conflicts of interest.
References
- Ara R, Reynolds A, Conway P. The cost-effectiveness of etanercept in patients with severe ankylosing spondylitis in the UK. Rheumatology (2007) 46:1338–44.
[Abstract/Free Full Text] - Botteman MF, Hay JW, Luo MP, Curry AS, Wong RL, van Hout BA. Cost effectiveness of adalimumab for the treatment of ankylosing spondylitis in the United Kingdom. Rheumatology (2007) 46:1320–28.
[Abstract/Free Full Text] - Kobelt G, Andlin-Sobocki P, Brophy S, Jönsonn L, Calin A, Braun J. The burden of ankylosing spondylitis and the cost effectiveness of treatment with infliximab (Remicade). Rheumatology (2004) 43:158–66.
- Kobelt G, Sobocki P, Sieper J, Braun J. Comparison of the cost effectiveness of infliximab in the treatment of ankylosing spondylitis in the United Kingdom based on two different clinical trials. Int J Technol Assess Health Care (2007) 23:368–75.[CrossRef][ISI][Medline]
- McLeod C, Bagust A, Boland A, et al. Adalimumab, etanercept, and infliximab for the treatment of ankylosing spondylitis: a systematic review & economic evaluation. Health Technol Assess (2007) (8 August 2007, date last accessed). 11. http://guidance.nice.org.uk/page.aspx?o=394626.
- Bansback N, Wailoo A. Comparisons of Submitted Models for the Appraisal of Etanercept, Adalimumab and Infliximab for Ankylosing Spondylitis. Report by the Nice Decision Support Unit (2007) (8 August 2007, date last accessed). Available at http://guidance.nice.org.uk/page.aspx?o=441923.
- Wailoo A, Bansback N, Chilcott J. Ankylosing Spondylitis: a Comparison of Cost Effectiveness Models. Report by the Nice Decision Support Unit. (2007) (8 August 2007, date last accessed). Available at http://guidance.nice.org.uk/page.aspx?o=441926.
- Jois RN, Gaffney K, Keat A. Anti-tumour necrosis factor therapy for ankylosing spondylitis—unresolved issues. Rheumatology (2007) 46:899–901.
[Free Full Text] - Drummond MF, Jefferson TO. Guidelines for authors and peer reviewers of economic submissions to the BMJ. Brit Med J (1996) 313:275–83.
[Free Full Text] - Weinstein MC, O’Brien B, Hornberger J, et al. Principles of good practice of decision analytic modeling in health care evaluation: report of the ISPOR Task Force on Good Research Practices-Modeling Studies. Value Health (2003) 6:9–17.[CrossRef][ISI][Medline]
- Bansback N, Maetzel A, Drummond M, et al. Considerations and preliminary proposals for defining a reference case for economic evaluations in ankylosing spondylitis. J Rheumatol (2007) 34:1178–83.[ISI][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||