C. Green
INTRODUCTION
In healthcare systems around the world there are limits on the funding available for healthcare and there is a growing demand for a broadening range of health services. These factors create an environment where healthcare systems are unable to provide all healthcare inter- ventions that are known to be clinically effective and potentially beneficial to respective patient groups. This gives rise to an increasing number of difficult decisions, and increasing pressure is placed on those involved in decision-making within healthcare. Nevertheless, decisions have to be made. The focus of pharmacoeconomics is on making the most of avail- able healthcare resources, and on the provision of information to help with difficult health policy decisions (e.g. which health services should be available, and to whom).
Pharmacoeconomics considers the relative value of specific health interventions using the framework of economic evaluation. Economic evaluation is the comparative analysis of alter- native courses of action in terms of both their costs and consequences [1]. It aims to promote efficiency, or to use more familiar terminology, it considers the issue of ‘value for money’.
There are a number of forms of economic evaluation (cost minimization analysis, cost- effectiveness analysis, cost utility analysis), but they have the common feature that some combin- ation of inputs to a healthcare service or programme are compared with some combination of outputs. Economic evaluation presents information on the cost-effectiveness of alterna- tive strategies – with efficiency often one of a number of considerations relevant to decision makers. This form of information is now prominent in the world of health policy. In the UK, the explicit use of economic evaluation and cost-effectiveness analysis to guide health pol- icy decisions is highlighted in the health technology appraisal process of the National Institute for Health and Clinical Excellence (NICE) [2], and the guidance published by NICE. However, there are a great number of other decision-making bodies that take account of cost-effectiveness when making health policy decisions at what may be a local, regional, national or international level.
In order to offer some understanding of pharmacoeconomics in the area of dementia, I offer a commentary on the economic evaluation of drugs for Alzheimer’s disease (AD). I present a broad review of the literature on the cost-effectiveness of (i) acetylcholinesterase inhibitors for the treatment of mild to moderately-severe AD, and (ii) memantine for the treatment of moderately-severe to severe disease. The chapter begins with an outline sum- mary of the literature to inform on the cost-effectiveness of these drugs for AD, followed by a critical summary of the published studies. This summary draws out some important issues and concerns when interpreting the literature.
Colin Green, BA (Hons), MSc, Principal Research Fellow, Southampton Health Technology Assessments Centre (SHTAC), Wessex Institute for Health Research and Development, University of Southampton, Southampton, UK
©Atlas Medical Publishing Ltd 2007
PHARMACOECONOMIC STUDIES (COST-EFFECTIVENESS ANALYSES) IN ALZHEIMER’S DISEASE
This summary review is based on a detailed systematic review of the literature undertaken to inform decision makers in the UK National Health Service (NHS). The methods for the literature review have been reported in detail elsewhere [3]. The search strategy considered literature available from inception of databases (e.g. MEDLINE, EMBase) up to mid-2004. It identified a growing literature (with the majority of studies published since 2000), with 18 published economic evaluations, six published abstracts, and two published UK NHS regional reports. There were also a number of review papers, either on individual drugs [4–7] or providing a broader review on one or more of these drugs [8, 9]. The current dis- cussion is on the published studies, and review papers have not been covered. Interested readers will find these review papers accessible and helpful in combination with this chap- ter if they have a keen interest in the cost-effectiveness literature.
Table 7.1 provides a summary of the scope of the literature, the headline messages and the basic characteristics of the published literature. The majority of studies are on the assess- ment of donepezil, and this reflects the fact that donepezil was the first of these products to be available for the treatment of AD. Of some concern, is that at least 16 of these cost-effectiveness studies are directly or indirectly supported by the pharmaceutical industry (manufacturers), and may therefore, but not necessarily, have biases due to competing interests. All studies are drug-specific, with single drug comparisons to placebo (usual care), and they all present country-specific analysis. All except two of the cost-effectiveness studies have used a model as the analytical framework for analysis. Models, using clinical, economic and epidemio- logical data, are used to consider AD progression over time. Studies have relied on model- ling due to the absence of longer-term effectiveness data, with clinical trials rarely extending beyond 48 weeks. Given that available effectiveness data is on specific intermediate mark- ers of disease progression (e.g. scores on cognitive function) and that there is an absence of longer-term data on patient related outcomes (e.g. reduced rates of institutionalization, reduced need for full-time care, and quality of life gains), some form of modelling has been unavoidable. Studies have extrapolated from short-term trial data to longer-term outcomes, using a number of different modelling approaches, and these are discussed below.
Perspective (viewpoint) of the analysis is an important issue for those wishing to draw some conclusions from the literature. In at least 10 studies the perspective is that of a soci- etal decision maker, and in much of the remainder the perspective is often unclear.
Almost all studies report that drugs offer health benefits and cost savings over time, indi- cating that their use is cost-effective. Yet, this may not be the case when considering treat- ment under certain conditions, for example when using a third party payer perspective (e.g.
UK NHS). One study [10] reports that treatment incurs added costs, but indicates that it is a cost-effective use of resources, whilst one study [11] indicates cost savings when not including the cost for the drug used. A recent UK study [12] reports that donepezil vs.
placebo results in a reduction in cognitive decline, but that donepezil is not a cost-effective use of UK NHS resources.
CRITICAL APPRAISAL OF ECONOMIC EVALUATIONS
The basic tasks of any economic evaluation are to identify, measure, value and compare the costs and consequences of the alternatives being considered. There are always challenges in the conduct of economic evaluation and studies can vary greatly in the methods they use and in their quality. Therefore, it is important to critically appraise the cost-effectiveness lit- erature available when it is being used to inform a particular policy question. Any user of cost-effectiveness data should take note of the methods used and the generalizability of the data, as well as their results. A decision maker should have in their mind the important
acoeconomic studies75 Results presented
Country/ Base year Modelling Cost saving or Benefits: delays to Competing References Drug setting Perspective costs/prices studies cost neutral disease progression interests*
Stewart et al. [21] Donepezil UK NS/societal 1996/1997
Jonsson et al. [22] Sweden NS/societal 1995?
O’Brien et al. [23] Canada Societal 1997
Neumann et al. [16] USA Societal 1997
Ikeda et al. [24] Japan TPP 2000
Fagnani et al. [25] France Societal 2003
Wimo et al. [17] Sweden Societal 1999 X Trial
based
AD Collaborative UK Societal 2002/3 X Trial Added costs None
Group [12] based
Fenn and Gray [11] Rivastigmine UK UK NHS and PSS 1997
Hauber et al. [43] USA NS/unclear 1997
Hauber et al. [38] Canada Societal 1997
Getsios et al. [58] Galantamine Canada NS/TPP 1999
Garfield et al. [44] Sweden NS/TPP 1998
Caro et al. [45] Netherlands Part societal 1998
Migliaccio-Walle USA Third party payer 2000
et al. [46]
Ward et al. [10] UK UK NHS 2001 Added costs
François et al. [30] Memantine Finland Societal 2001
Jones et al. [19] UK NS/TPP 2003
TPP⫽third party payer; NS⫽not stated; NS/societal⫽not stated but appears societal; NS/TPP⫽not stated but appears third party payer.
*Manufacturer involvement.
elements of a sound economic evaluation when taking account of whether the cost-effective- ness analyses available are useful in their particular setting. There are a number of pub- lished sources of methodological standards for economic evaluation [1, 13], and there is useful guidance on the elements of good practice in cost-effectiveness modelling studies [14]. Detail on these methodological standards will not be offered here; that job has been done well by others elsewhere [1, 14, 15]. But one of the objectives of this paper is to outline the basic methods underlying economic evaluation, therefore Table 7.2 presents a summary of the important elements of an economic evaluation, which are often summarized in the form of a checklist of questions to ask about a published study [1, 13].
Table 7.3 reports basic data from a critical appraisal of the published economic evalu- ations. The framework for critical appraisal used is adapted from that presented by Drummond et al. [1], and in this instance – for presentational purposes only – some simple and crude symbols have been used to draw attention to those areas of the literature that are deserving of particular attention. Ticks are used to indicate a favourable (acceptable) response to a review area, crosses to indicate a non-favourable response, and question marks to indicate issues which are unknown or uncertain. I would not advocate this simple approach generally, and would recommend that when undertaking a critical review of this nature it is advisable to use the checklist as a prompt, and to note in a descriptive manner the findings from studies. A more detailed account of the review can be found elsewhere [3].
However, Table 7.3 is useful in drawing attention to the areas of particular concern within the published cost-effectiveness studies. Whilst there may be some concerns with the style of reporting and presentation of analysis (e.g. stated study question, use of incremental analysis, reporting of sensitivity analysis) these areas are not a major concern in the context of the current discussion. Those areas that are of greater concern, and form the basis of our discussion of the literature, are (a) analytical perspective, (b) the reporting and use of effect- iveness data, (c) the methods used to model progression of disease, and (d) the related issues of identification, measurement and valuation of relevant costs and consequences.
THE IMPORTANCE OF PERSPECTIVE/VIEWPOINT
The perspective, or viewpoint, of an economic evaluation determines the scope of the analy- sis, i.e. which combination of inputs and outputs are relevant. The analytical perspective of a study can be that of the individual patient, a specific institution, a target group for specific services, the national health budget, the overall public sector budget, or a broad societal per- spective. Indeed a study can address any or all of these perspectives. With perspective, it is not a simple question of which perspective an economic evaluation should take, as an eco- nomic evaluation can aim to inform against a number of different objectives and perspec- tives. Any analysis should be clear about the perspective taken, or the multiple perspectives
1. Was a well-defined question posed in answerable form?
2. Was a comprehensive description of the competing alternatives given?
3. Was the effectiveness of the programmes or services established?
4. Were all the important and relevant costs and consequences for each alternative identified?
5. Were costs and consequences measured accurately in appropriate physical units?
6. Were costs and consequences valued credibly?
7. Were costs and consequences adjusted for differential timing?
8. Was an incremental analysis of costs and consequences of alternatives performed?
9. Was allowance made for uncertainty in the estimates of costs and consequences?
10. Did the presentation and discussion of study results include all issues of concern to users?
Table 7.2 Checklist for assessing economic evaluations (adapted with permission from Drummond et al. [1])
acoeconomic studies77 Effect Relevant costs and consequences
related to Differential Incremental Sensitivity Modelling
Question? Effectiveness population Measured Valued timing analysis analysis conducted Study Alternatives? established? of interest? Identified? accurately? credibly? considered? undertaken? undertaken? reasonably?
Donepezil
Stewart X X/? ? ? ? ?
et al. [21]
Jonsson /? X/? /? ? ? ?
et al. [22]
O’Brien /? ? ?
et al. [23]
Neumann X ? ? ?
et al. [16]
Ikeda X ?/X ? /X ?
et al. [24]
Fagnani X /? ? /? ? ?
et al. [25]
Wimo N/A X N/A
et al. [17]
AD Collaborative X N/A
Group [12]
Rivastigmine
Fenn and ?/ ? /? ? ? X X ?
Gray [11]
Hauber /? X /? ? ? ? X ?
et al. [43]
Hauber X /? /? ? ? X ?
et al. [38]
Therapeutic Strategies in Dementia
Getsios ? /? ? ? ?/
et al. [58]
Garfield X X /? ? ? /? ?/
et al. [44]
Caro X X /? ? ? ?/
et al. [45]
Migliaccio- X ? /? ? ? ?/
Walle et al.
[46]
Ward X X /? /? /? ?/
et al. [10]
Memantine
François X X ? /? /? ?/
et al. [30]
Jones X X ? ?/X ?/X ?/
et al. [19]
Reviewer (CG) opinion: ⫽judged OK; X⫽judged not OK or insufficient information reported; ?⫽unknown/uncertain; /?⫽judged potentially OK but uncertain; X/?⫽judged to be not OK, but uncertain.
presented, and the component parts of the cost-effectiveness calculation, i.e. which costs and consequences have been taken into account. A health intervention which looks attract- ive from one perspective may look much less attractive when other perspectives are con- sidered. This can be the case for AD. AD is not a simple condition for evaluation. Treatment for AD involves a number of different provider and funding sectors, and patients and carers often contribute to the cost of care. Informal care (often family members, but also volun- teers) is one of the most important resources in AD care, but different perspectives will dictate whether informal care forms part of the economic analysis. Drug treatments are relatively low cost interventions, in the context of broader care costs, but still remain a significant budgetary concern for healthcare providers, and will be very important from a third party payer perspective. A societal perspective will include all relevant costs and benefits regardless of where they occur and regardless of who pays. A specific payer perspective will limit the economic analysis to take into account only the costs that fall on the budget of that payer (i.e. exclude patient and carer costs). However, even where a deci- sion is taken to include societal costs (e.g. informal care costs) there are a great number of difficulties estimating and valuing resource inputs and benefits from a societal perspective, and these remain methodological issues that have yet to be resolved.
When reviewing the literature, in a number of cases the perspective is not clearly stated, and often when it is stated there remain uncertainties over the actual scope of the input data used. Table 7.1 reports summary detail on perspective in published studies, with six of the 18 studies not stating the perspective of their analysis. Where possible the perspective that is apparently taken has been indicated (e.g. societal or third party payer), but this is not always possible. Even where a perspective is clearly stated it is important to consider if that perspective has been pursued correctly in the analysis undertaken. It is clear from the review undertaken that there are some ambiguities and/or uncertainties when interpreting the literature in the context of stated perspective. For example, where societal perspective is stated by a number of studies [16, 17], all of the consequences of treatment may not have been investigated (e.g. the effect of treatment on carers is not included, other than in the esti- mation of longer-term cost consequences). While six studies state, or indicate, that their per- spective is that of a third party payer (e.g. NHS and personal social services in the UK), we can easily conclude that certain costs are included in the analysis which are not met by the third party payer. For example, in UK studies [10, 11], costs that are not met by the payer (e.g. UK NHS) are included in the analysis. In the UK the costs for longer-term care (i.e.
institutional costs) are not met by the UK or public sector budget for all patients. Support for such costs is dependent on the financial status of the patient (means tested) and around 30% of those patients with AD in a UK long-term institutional setting are responsible for their own care costs [18]. The UK cost-effectiveness analysis for memantine, presented by Jones et al. [19] as a reflection of a third party payer perspective (although this is not directly stated) includes estimates of longer-term care costs that far exceed the usual (or typical) level of funding provided by the UK NHS and personal social services.
METHODS USED TO MODEL DISEASE PROGRESSION AND TREATMENT EFFECT
It is now well accepted that in economic evaluation we often require a model of the disease and its management. It is rare that a single clinical trial will provide all the necessary infor- mation on costs and outcomes, and even if this were to be the case it would still be neces- sary to model the available data in the setting and population of interest (outside of clinical trial protocol). Indeed there has been an acceptance that models and trials are not alterna- tive analytical frameworks, and that they are both necessary components of the evaluation process. AD is a good example of the necessity to use a model to synthesize available clini- cal trial data with data on costs, epidemiology and other data on disease management, in order to predict longer-term outcomes and costs. In the studies identified to inform on the
cost-effectiveness of drugs for AD, we see a variety of methods used to model disease progression and the impact of treatment. In each of the four drugs of interest, we see a dif- ferent approach to the modelling of cost-effectiveness data. For the three cholinesterase inhibitors this makes the comparison of the drugs more difficult, and does not allow the reader to consider how each of the drugs would compare when using a common analytical framework and similar model parameter inputs. Memantine is available for the treatment of more severe disease, therefore it is understandable that a different modelling approach is adopted. In all studies using cost-effectiveness models there are concerns related to struc- ture and/or data used. I offer a brief summary of the methods applied for each of the four drugs, and highlight those areas of greatest concern.
Economic models have typically been structured and analysed using decision trees or Markov models. Markov models represent the course of a disease in terms of mutually exclusive ‘health states’ and the transitions among them [20]. Markov models are particu- larly useful when a decision problem involves clinical changes that are ongoing over time.
The progression of AD with time is a good clinical example.
The cost-effectiveness studies for donepezil are almost entirely based on the use of Markov type models of disease progression. Health states defined according to categories of cognitive function have been used to estimate disease progression across different levels of disease severity, with treatment effects based on trial-specific data. Three studies have used Mini-mental State Examination (MMSE) scores to define either four or five levels of AD severity [21–23], two studies used Clinical Dementia Rating (CDR) scale scores to define three levels of disease severity [16, 24], and one study [25] used MMSE in a continuous man- ner. Progression of disease is modelled using transition probabilities between health states at each model cycle (e.g. 6-month cycle), with an ongoing risk of death over time included in the models. There are variations between studies in the methods used to determine appropriate transition probabilities for the disease progression models. Three studies obtained transition probabilities for donepezil-treated and untreated groups from clinical trial data [22, 23, 25]. One study [21] used epidemiological data for the untreated control group and trial data for the donepezil-treated group, while the others used epidemiological data to calculate transition probabilities for the untreated group and then applied a risk reduction factor derived from clinical trial data to generate transition probabilities for the treatment group [16, 22, 24]. In some studies [21, 23, 25] the effect of donepezil on disease progression was assumed to last for only part of the overall time horizon, but other studies [16, 22, 24] assumed that the treatment effect persisted for the entire time horizon. Generally models incorporated an ongoing mortality risk that was the same for both treated and untreated patients; this mortality risk was dependent on disease severity in three studies [16, 22, 24]. Cycle length, time horizon, and the characteristics of the baseline patient cohorts varied across studies. All the studies assumed that donepezil treatment would stop when patients reached a state of severe AD.
The studies informing on the cost-effectiveness of rivastigmine have all used the same modelling framework. They have used the hazard model of disease progression presented by Fenn and Gray [11]. This model uses individual patient data from clinical trials on rivastigmine [26, 27] to estimate the time taken for each patient to move from one level of AD severity to another. The model estimates the likelihood that a patient will remain at a particular MMSE score at any given time. Statistical techniques are used to model disease progression. The model is used to generate survival curves for both placebo and rivastig- mine treatment groups, with these extrapolated beyond the end of the trial period. The impact of treatment on disease progression is measured as days saved by preventing patients from entering the next, more severe, stage of AD. This delay in disease progression is represented by the area between the placebo and treatment survival curves. The hazard model does not incorporate a mortality risk directly in the disease progression process. The proportional difference in cognitive decline between the placebo and treatment groups