This section is intended to provide a general view of the pharmacoeconomic methods utilized in this thesis. The specific methods used in the studies will be described in detail under their own chapters. Markov modeling (Briggs and Sculpher 1998) was used in three of the studies (II, III, V). All of these used a model with 3 mutually exclusive health states (Figure 9). Markov model stages are characterized as discrete, mutually exclusive, and they have no memory (Kuntz and Weinstein 2001). In addition, all patients within a stage are assumed to be homogenous. Time in Markov models is applied as cycles, and a hypothetical patient population transitions between the predetermined health stages following these model cycles (Briggs and Sculpher 1998, Kuntz and Weinstein 2001). The health stages and cycle length are determined according to the study perspective and disease characteristics.
The model structure and transition probabilities are intended to illustrate the natural flow of the disease concerned. Since all diseases are unique, the same evaluation model can rarely be used for different health conditions as such. Thus, the 3-stage model (Figure 9) was adjusted to fit the aim and purpose of the individual studies. The technical realization of the model varied between the studies, although the basic structure of the model remained the same. The structure of the framework model, including health states “No progression”, “Progressed disease” and “Dead” is governed by the assumption that these health states apply to most types of cancer. This partition is present in clinical trials showing end-points such as time to treatment failure, time to disease progression, progression-free survival and overall survival. The utilized partition concerning disease severity is also present in terms of treatment costs. When treatment is modified from active to supportive/palliative, the cost structure changes from drug-intensive to being hospital-intensive (study III).
Figure 9. Basic framework for the utilized cancer model
Probabilistic sensitivity analyses (PSA) were utilized in both of the cost-effectiveness analyses (II, V), and in the budget impact analysis of adjuvant trastuzumab (IV). In PSA, each of the chosen model inputs is allowed to vary independently according to the predetermined probability distributions, in order to incorporate the uncertainty related to the model parameters (Briggs 2001). In study II, the model was built with WinBUGS software, which directly implements a Bayesian approach (Martikainen 2008). The PSA results were depicted as cost-effectiveness planes (V), cost-effectiveness acceptability curves (II, V), and affordability curves (IV). The affordability curve in BIA shows the probability of staying within a given budget (Sendi and Briggs 2001).
Value of Information analysis (VOI) was applied to the probabilistic model in study V.
The value of additional research informs what are the maximum costs that one would be willing to pay to reduce uncertainty. This may also be used to quantify parameter uncertainty, in order to more effectively allocate research resources. (Barton et al. 2008)
It is relevant to take population dynamics into consideration in budget impact models with chronic diseases, unstable populations, and in studies using a long time horizon. In this thesis the population dynamics was handled through state transition models (Kuntz and Weinstein 2001). The two cost-effectiveness models (II, V) utilized closed cohort modeling, where the hypothetical patients were followed through their lifetime. Both of the budget impact models (III, IV) were open cohort stage-transition models, where new patients entered the model according to the estimated incidence of new cases. New cases entered the patient pool and stayed there until the end of follow-up or death. Treatment effectiveness was incorporated into the budget impact models. This enabled accounting for differences in costs within different disease stages between the treatments being compared.
A general illustration of patient dynamics utilized in the open cohort models in this thesis is depicted in Figure 10.
Figure 10. Illustration of patient dynamics in the utilized budget impact models
Economic evaluations may be performed from several perspectives (Figure 11), which determines the included resource use and costs. In cost-effectiveness analyses, the most commonly the utilized perspective is that of society or of the health care payer. The payer perspective is considered to be most relevant for budget impact analyses (Annemans 2010), while the use of societal perspective is rare but cannot be excluded. The study perspective derives from the aims of the study, and thus it varies among the studies included in this thesis.
Figure 11. Different perspectives of economic evaluations (Mogyorosy and Smith 2005)
Irrespective of the chosen model structure, a model itself will produce nothing without proper inputs. It may be said that a model is only as good as the inputs it contains. Thus, the selection of reliable data sources is a crucial step in the modeling process. Figure 12 illustrates the type of model inputs that are required in cost-effectiveness and budget impact models. Decisions on model inputs are based on the requirements of a study and the availability of data. Thus, the model inputs presented in Figure 12 may not completely apply to all circumstances, as some disease-specific or treatment-specific features may be lacking. The choice of data sources and model inputs is often based on individual decisions.
Nevertheless, guidelines for measuring drug costs have been presented by the International Society for Pharmacoeconomics and Outcomes Research (ISPOR). In addition, several recommendations for good practice in different modeling techniques are currently under preparation by ISPOR.
Figure 12. Required model inputs in cost-effectiveness and budget impact analyses. The dotted lines represent optional inputs depending on study perspective.
5 Specific aspects and methodological challenges of pharmacoeconomics in cancer care
Pharmaceuticals go through a rigorous evaluation process prior to obtaining marketing authorization and a possible reimbursement status. Randomised clinical trials are considered as the golden standard for demontrating the efficacy of new therapies. The regulatory authorities, such as European Medicines Agency, use these data along with other additional information to assess the balance between risks and benefits of the intervention. The regulatory approval process, however, does not consider issues related to costs or funding of the treatment. In the European Union, the costs and reimbursement policies are considered within each of the member states. (Tannock et al. 2011) In Finland, health economic evaluations are required for medicines with a new active pharmaceutical ingredient, when applying for a reasonable wholesale price and reimbursement status (Laki sairausvakuutuslain muuttamisesta 802/2008). Nevertheless, economic aspects are only one of the issues that are considered when funding decisions are made, since these are also affected by clinical issues and equity. Other issues influencing these decisions are the degree of uncertainty related to the results, the innovative nature of the technology, certain features of the disease, characteristics of the target population, and issues concerning the wider societal costs and benefits (Barry 2007).
Figure 13 presents a schematic illustration of the requirement that a pharmaceutical faces when entering the Finnish market. Marketing authorization is rarely sufficient for a successful market entry, especially among prescription products that are used in outpatient care. If a medication is not considered to be suitable for reimbursement, then its market penetration will prove difficult. Successful entry to the market most often requires demonstrations of cost-effectiveness and affordability (i.e. ability to pay for the treatment).
Figure 13. Pharmaceuticals entry to Finnish market via the European centralized marketing approval process
5.1 LIMITATIONS RELATED TO CLINICAL TRIALS AS A DATA SOURCE