• No results found

The cost-effectiveness of proton radiation therapy: a case study at the UMCG

N/A
N/A
Protected

Academic year: 2021

Share "The cost-effectiveness of proton radiation therapy: a case study at the UMCG"

Copied!
65
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

The cost-effectiveness of proton radiation therapy: a case study

at the UMCG

Rutger Mulder Student number: s2051370 MSc Finance & MSc BA-O&MC Faculty of Economics and Business

Groningen University Supervisor MSc BA-O&MC Prof. Dr. D.M. Swagerman Faculty of Economics and Business

Groningen University Supervisor MSc Finance

Dr. W. Westerman

Faculty of Economics and Business Groningen University

Abstract

The purpose of this paper was to guide future cost-effectiveness research by the construction of a comprehensive conceptual model. In addition, the University Medical Centre Groningen (UMCG) was used as case study object to explore the cost-effectiveness of proton radiation therapy from a health care insurance perspective. In doing this the current study constitutes a valuable contribution to medical cost-effectiveness research in general and radiotherapy cost-effectiveness research in particular. An altered version of the simulation model of Quik et al. (2014) was used to examine and compare costs and Quality-Adjusted Life Years (QALYs) of nine patients. Results showed us that proton radiation therapy is regarded as the cost-effective treatment for 33% of selected patients. Another 33% of patients is on the edge of being regarded as cost-effective. On the one hand these results are very promising since they prove the relevance of proton therapy. On the other hand however, the results reveal the weakness of current selection criteria for proton therapy. It is therefore recommended that selection criteria should be tightened. The combination of high demand and scarce supply legitimize stricter rules for selecting patients without experiencing oversupply.

Keywords: proton therapy, photon therapy, case study, cost-effectiveness, conceptual model

JEL classifications: I13, I19

(2)

1

Preface

The master thesis that is lying in front of you is the result of half a year research at the radiotherapy department at the University Medical Centre Groningen (UMCG). Other departments involved in this study are: board of directors, epidemiology, and special dentistry. This study was initiated as a response to the public debate around the introduction of proton radiation therapy in the Netherlands. On the one hand, the Dutch UMCs strive to provide state of the art health care. On the other hand, insurance companies committed themselves to the duty of providing affordable health care to all citizens, meaning they critically review the cost-effectiveness of health care innovations. In a democratic society, discussions and debates like this are inevitable and often lead to a consensus that satisfies the majority. Research has shown to be essential in this consensus making process.

(3)

2

Table of contents

Preface 1 Table of contents 2 Introduction 4 2. Problem definition 7 2.1 Background 7

2.2 Physical, clinical, and cost advantages 8

2.3 Research questions 9

3. Literature review 11

3.1 Cost drivers 11

3.1.1 Direct health care costs – capital and operational expenditures 11 3.1.2 Indirect health care costs – future costs from complications 12 3.1.3 Direct non-health care costs – travel and hotel costs 15 3.1.4 Indirect non-health care costs – changes in productivity 15

3.2 Capital budgeting 16

3.2.1 Non-DCF methods 16

3.2.2 DCF methods 17

3.2.3 Alternative methods used in capital budgeting 19

3.3 DCF input 20

3.3.1 Cash flows 21

3.3.2 Cost of capital 22

3.4 Conceptual model 24

4. Data and methodology 26

(4)

3 5.1 Model input 32 5.1.1 Cost input 32 5.1.2 HRQoL input 34 5.1.3 NPV model 35 5.2 Outcomes 35 5.2.1 Standard case 35 5.2.2 Sensitivity analyses 37 6. Discussion 38 7. Conclusion 41 References 42

Tables & figures 47

Appendix A 54

Appendix B 56

Appendix C 59

(5)

4

Introduction

The UMCG represents itself as a knowledge institution, hospital, education and training institution, and innovation centre. These pillars are a necessity to comply with their ambitious mission: ‘Building the future of health care’. As a public institution the UMCG aims to be of added value to society. This added value will in the upcoming years, like in the past, be expressed through the focus on Healthy Ageing. To achieve this, the UMCG is cooperating with all faculties of the University of Groningen in order to integrate knowledge from different schools of thought. In this way the UMCG wants to compete with the best bio-medical knowledge institutions from around the world (Annual report UMCG, 2013).

As one of the most progressive and innovative UMCs of the Netherlands, the UMCG is busy setting up a hospital-based facility – the UMC Groningen Proton Therapy Centre – as an integral part of the UMC Groningen Cancer Centre. The GPTC will be the first proton facility in the Netherlands. The GPTC strives to bundle care, research, education, and technological innovation to accelerate the introduction of proton therapy (Groningen Proton Therapy Centre, www.umcgroningenptc.nl). The Groningen facility will be unique in combining treatment with an extensive Research & Development (R&D) programme to demonstrate the clinical and economic benefits and to continuously improve proton therapy technology and the treatment itself. This R&D programme is called PARticle Therapy REsearch Centre (PARTREC). The main aim of PARTREC is to improve treatment quality at reduced costs by both preventing damage to healthy tissue and by improving tumour control.

(6)

5

Insert figure 1 about here

Already back in 1946, Robert Wilson hypothesized that the greater precision of proton therapy enables a reduction of radiation dose to healthy tissue and thus the probability and seriousness of complications, while keeping to radiation dose to the tumour constant. Or reasoned the other way around, with an equal dose of radiation to healthy tissue the tumour will receive more radiation. The former will result in fewer complications since more healthy tissue is saved, while the latter results in better tumour control and reduced change of recurrence. Shortly after Wilson’s discovery scientists initiated the first studies on proton therapy to confirm this hypothesis. It was however not until the John M. Slater Proton Therapy and Research Centre became operational in 1990 that the full benefits of proton therapy were available to cancer patients. Despite knowing it for almost seven decades proton therapy is still in its infancy stages. At this very moment the number of proton facilities is in the low thirties and the new facilities which may be built in the next five years are comparable in number (Proton Therapy Centre Switzerland, www.ptcs.ch).

(7)

6

allowed me to simulate life course of patients having received photon radiation. Economic evaluations in health care differ from economic evaluations in business because benefits are measured in terms of health care effects rather than money. Therefore a cost-effectiveness study seems most appropriate in this setting since it allows us to evaluate whether the additional health care effects of a new therapy outweigh its additional costs.

(8)

7

2. Problem definition

This chapter elaborates on the reasons that form the basis of this study. The first paragraph discusses the background of proton therapy in the Netherlands. This background is essential since it explains the need for Dutch society to find out the cost-effectiveness of this therapy. The next paragraph discusses the steps that need to be taken in order to estimate the cost-effectiveness. The third paragraph formulates the objective of research and its corresponding research questions.

2.1 Background

The National Health Care Institute, advises the government on insured health care. The board recommended the ministry of Health, Welfare, and Sport that proton radiation therapy should become insured care for Dutch citizens with standard or model-based indications. The former includes: intra-ocular, skull base- and paraspinal, and pediatric tumours and the latter includes: head and neck, lung, prostate, and breast tumours. Subsequently the Ministry released licences to four UMCs in the Netherlands: Groningen, Amsterdam, Maastricht, and Delft to treat patients with proton radiation therapy. This announcement however initiated an enormous public debate between the UMCs, insurance companies, the ministry, and the Dutch care authority. Most resistance originates from insurance companies, they argue that four licences is not a controlled introduction of proton therapy (Mosca, 2014). In addition they state that releasing four licences in such a small country is in conflict with a recently launched programme by the Dutch government ‘reduction of waste in health care’. Furthermore they question the demand for proton therapy, if capacity is larger than demand, health care insurers fear supply-induced demand like in the US (Epstein, 2012). In other words, Epstein argued that patients will be treated with proton therapy because of its availability and not because of its necessity. Finally, they argue that societal interests should be taken into account to make sure future health care remains affordable.

(9)

8

standard and model-based indications and found 5,000 patients eligible for proton therapy in 2005, meaning approximately 8,000 in 2020 (Langendijk et al., 2012). Based on these findings of the two advisory bodies of the Dutch government, a capacity of 2,200 patients can be seen as a gradual and controlled introduction of the new treatment. Patient demand for proton therapy does not seem to be a problem in the near future. The second argument from the insurance companies regarding affordable health care costs seems however a bigger problem. In order to achieve mutual agreement among all stakeholders further research need to be done to examine whether the benefits (i.e. additional health care effects) of new technologies are worth the extra cost.

2.2 Physical, clinical and cost advantages

(10)

9

Despite of these encouraging literature reviews, literature is dispersed and inconclusive on the clinical advantages. Olsen et al. (2007) for instance mentioned that evidence on clinical efficacy of protons relies to a large extent on non-controlled studies, and is thus associated with low levels of reliability. They add that reported studies are heterogeneous in design and do not allow for comparison. Hall and Phill (2006) discuss an example that leads to incomparability. They state that only proton facilities that use a pencil scanning beam show clinical advantages compared to photon therapy. Miller et al. (2013) argue that randomized controlled trials remain the ideal tool for research in proton radiotherapy. Lievens and Pijls-Johannesma (2013) add that the main cause of bias in economic evaluations in proton therapy is the lack of valid data on effects as well as costs. They propose that proton centres develop prospective registries with the goal of long-term data collection to support currently available evidence. However, in the absence of level one evidence (i.e. randomized controlled trials), well performed modelling studies including available cost and effect parameters can help to address this problem of invalid data (Lievens and Pijls-Johannesma, 2013). In so called ‘in silico planning comparative’ (ISPC) studies, proton and photon treatment plans are compared. Based on the dose distributions to organs at risk (OAR), complication probabilities can be estimated by making use of existing Normal Tissue Complication Probability (NTCP) models.

The last link required to estimate the cost-effectiveness of proton therapy involves the quantification of clinical advantages in terms of costs and effects. Previous literature only contains four attempts to determine the cost-effectiveness of proton therapy, all with their weaknesses (Lundkvist et al., 2005 [1]; Lundkvist et al., 2005 [2]; Ramaekers et al., 2012; Quik et al., 2014). To make solid decisions on whether this new therapy should be implemented or not, mutual agreement should be achieved. Therefore it is essential to continue research on the cost-effectiveness of proton therapy.

2.3 Research questions

(11)

10

therefore taken from a societal perspective. The research questions in this section steer the literature review of chapter three which in its turn forms the basis of the conceptual model. The main research question of this paper is:

Is proton radiation therapy cost-effective when compared to currently used photon radiation therapy?

Cost-effectiveness is however a broad term which cannot be measured at once. Therefore several sub-questions were constructed in a deductive manner in order to get a better grasp of all concepts that need to be addressed. The first step involves the identification of cost drivers. To construct a comprehensive model including all societal costs, direct as well as indirect cost drivers should be identified. The first sub-question reads:

Which direct and indirect cost drivers are relevant in evaluating the cost-effectiveness of a new medical therapy from a societal perspective?

Since the decision to adopt a new medical therapy is actually an investment decision, capital budgeting literature is addressed. Capital budgeting literature discusses many different techniques used to assist in making investment decisions. The third sub-question reads:

Which capital budgeting methods can be used in performing cost-effectiveness evaluations?

As touched upon before, this study compares costs as well as effects. To do this, an incremental cost-effectiveness ratio (ICER) should be calculated by dividing the change in costs (∆ costs) by the change in effects (∆ effects). The recognition of indirect future costs caused by a medical intervention makes comparison between two therapies a challenging task. Therefore discounting of cash flows (i.e. health care costs and health care effects) is required in order to compare two therapies at one point in time. Since health care economic evaluations are completely different from economic evaluations in business, special attention should be paid to the determination of cash flows and cost of capital. Both input variables changes with one’s perspective of analysis. This resulted in the fourth sub-question:

(12)

11

3. Literature review

This chapter discusses state of the art papers from business and medical literature. Guided by chapter two, publications in both fields are considered with respect to the case study institution. In exploring the literature the author took a societal perspective, the output of this review forms the basis of the comprehensive conceptual model constructed in the last section.

3.1 Cost drivers

Often a societal viewpoint is taken in cost-effectiveness studies. This concerns the costs to all parties affected by the disease (i.e. individuals themselves, friends and relatives, the health care sector, third-party payers, and employers). In practice, however, all relevant costs and effects may not always be identified, and some may be hard to measure (e.g. psychosocial costs of pain and suffering). In economic evaluations of medical therapies, costs are usually divided between direct and indirect costs. Previous economic evaluations in this field mainly focused on direct costs while omitting significant indirect costs, this was done either because they adopt another perspective of analysis (e.g. health care perspective) or because they lack data and knowledge. The following sections discuss relevant cost drivers for cost-effectiveness studies.

3.1.1 Direct health care costs – capital and operational expenditures

Proton therapy is a promising treatment modality for cancer, but it is also questioned whether the therapy is “too expensive to become true” (Lievens and van den Bogaert, 2005). The initial (capital) and ongoing (operational) expenditures are substantially higher compared to the conventional photon therapy. Peeters et al. (2011) and Goitein and Jermann (2003) conducted, to a great extent, similar studies in which they compare the costs, both initial and ongoing, of proton therapy relative to photon therapy. The cost ratio’s (proton therapy/photon therapy) for both, capital and operational expenditures, found by these authors are comparable and shown in table 1. Goitein and Jermann (2003) argue that, like all innovations, proton therapy has a much greater scope for improvement relative to photon therapy since the latter already went through several decades of development, improvements and cost reductions while the former did not. Another thing worth mentioning is that Goitein and Jermann (2003) recommend to use the ratio of costs rather than the absolute costs for comparison purposes since actual costs might fluctuate considerably between different countries and settings.

(13)

12

Assuming a lifecycle of 30 years for a proton facility and 15 years for a photon facility, the photon facility would have to replace equipment once in order to compare both therapies. This equipment replacement has been put into the capital expenditures budget of a photon facility. Total cost ratio’s including both, operational and capital expenditures, are 2.6 and 2.5 for respectively Peeters et al. (2010) and Goitein and Jermann (2003). Quik et al. (2014) ignored tariff calculations for both therapies and followed a recent institutional costing study, they estimated the costs of photon therapy in HNC at €17,500 versus €35,000 for proton therapy. The two other cost-effectiveness studies multiplied their own estimates of photon therapy with the cost ratio’s to find the costs of proton therapy.

3.1.2 Indirect health care costs – future costs from complications

Radiation-induced side-effects occur basically among all patients treated with radiotherapy. The seriousness of these side-effects is largely dependent on the dose distribution in the normal tissue. In particular irradiation of vital organs surrounding the tumour influence future health care costs (Ekwueme et al, 2014). Therefore previous cost-effectiveness studies recognized the need to include these costs in their evaluation (Lundkvist et al., 2005 [1]; Lundkvist et al., 2005 [2]; Ramaekerts et al, 2012; Quik et al., 2014). These studies developed simulation models to calculate the incidence of complications by combining ISPC studies and NTCP models. Implications of side-effects largely differ between patients and tumour types ranging from the need to take simple medication to need for complex medical interventions like surgery. Also, some side-effects might be temporarily where others are chronic. Based on the previously discussed higher precision of proton therapy it has become possible to develop treatment plans that spare healthy tissue and organs. This increased accuracy reduces the occurrence of radiation induced side-effects and therefore the need to use future health care services. If done consistently for all relevant complications of a specific tumour type, cost differences between both therapies will be reduced.

(14)

13

hypothyroidism. In addition Christensen et al. (2001) mention increased dental care costs as a direct result of xerostomia. Dental care costs however have seldom been examined before and might therefore be hard to quantify. Nonetheless this study is the first to include a rough estimate of radiation induced dental care costs.

Sticky saliva

Irradiation of glands can change the consistency of your saliva. It may become thicker, stringy and sticky, like mucus. Mucus doesn’t flow as well as normal saliva so it may build up in your mouth and throat. Simple medication is often prescribed to reduce the production of mucus. Sticky saliva has long been omitted in cost-effectiveness studies because of its low cost profile. Nevertheless, the most recent study of Quik et al. (2014) recommends to include sticky saliva in evaluating the cost-effectiveness. The NTCP model of Beetz et al. (2012) is used to estimate the incidence of sticky saliva at 6 months after treatment.

Xerostomia

(15)

14

Dysphagia

Radiation-induced xerostomia is said to be reduced by the use of new radiation therapies such as IMRT, therefore the problem of swallowing dysfunction is becoming one of the most relevant side-effects of radiotherapy. Several different grades of dysphagia are distinguished in literature. Patient complaints cannot always be classified into one of these grades, nevertheless they give some guidelines for diagnosis. Grade zero and one do not cause many difficulties because normal eating habits can roughly be preserved. Grade two requires the patient to eat pureed food. Grade 3 is however more problematic since mainly liquid food is taken, these liquid food bottles are expensive and are therefore a substantial cost driver for the health insurer. Dysphagia grade four or higher make the patient PEG or NEG tube dependent, this most severe state of dysphagia is discussed in the next paragraph. Again, both Ramaekers et al. (2012) and Quik et al. (2014) consider dysphagia and tube feeding as one of the most prominent side-effects of HNC radiation, and should therefore be included in cost-effectiveness research. The NTCP model of Christianen et al. (2012) is used to estimate the incidence of dysphagia two till four at six months after radiation.

Tube feeding

As already touched upon shortly, patients with dysphagia grade four or higher become completely PEG or NEG tube dependent. Percutaneous endoscopic gastronomy (PEG) and nasogastric (NEG) tubes are devices that patients with this severe grade of dysphagia use. The first being a tube directly inserted in the stomach and the second being a tube inserted through the nose. Tube feeding is by far the most expensive side-effect of radiation of HNC, first due to placing the tube in a surgical intervention and second due to daily consumption of two bags of liquid food. No single organ or dysfunction has been shown to determine overall swallowing because abnormalities are never experienced in isolation. However, radiation dose to pharyngeal muscles and larynx has proven to be of great influence on dysphagia (van der Laan et al., 2012). Therefore they are both included in the NTCP model of Wopken et al. (2014), this model is used in this study to estimate the 6 month incidence of tube feeding.

Hypothyroidism

(16)

15

tests need to be performed in order to check thyroid hormone levels to which medication doses can be adjusted. One out of ten patients need to see an endocrinologist once a year to receive some extra assistance. Despite the relatively low medical costs of this complication it is one of the most common side-effects and should therefore be taken into account (Quik et al., 2014). This complication is developed relatively late compared to the other complications, therefore the NTCP model of Boomsma et al. (2012) is used to estimate the 2 year incidence of hypothyroidism. According to Smith et al. (2009) the late onset rate was estimated at 4% a year and was therefore added to the NTCP.

3.1.3 Direct non-health care costs – travel and hotel costs

The public debate about the introduction of the proton therapy in the Netherlands is still going and therefore no final decision is made on the number of proton facilities. It seems however reasonable by now to assume the setup of at least one, maybe even two, proton facilities in the Netherlands. Considering other European countries we see that only large nations like Germany, France and Russia have more than one proton centre. The actual number of setups is important in evaluating the cost-effectiveness because it influences non-health care costs. The reasoning here is that transportation and accommodation costs will drop if more centres will be set up. Lundkvist et al. (2005) [1] and Lundkvist et al. (2005) [2] assumed that the majority of patients will be treated at a national proton facility outside of their home region. Based on population statistics and treatment praxis they assume that between 0%-70% of patients would have extra transportation and accommodation costs. In HNC patients, 35% was assumed to face on average €52.20 of daily accommodation costs and €217.40 of traveling costs. Lundkvist et al. (2005) [1] assumed additional costs of €883 for transportation and hotel during the treatment with proton therapy since patients would have to travel a longer distance to receive proton therapy compared to photon therapy.

3.1.4 Changes in productivity

(17)

16

since half of the cancer survivors are of working age. They estimated the per capita mean annual productivity loss for male survivors to be $3,719, compared with $2,260 among males without a cancer history. Female raised their annual per capita productivity loss from $2,703 to $4,033 due to cancer. In addition Fontenot et al. (2007) and Yoon et al. (2010) who showed a substantial reduction (26%-39%) in the risk of a secondary malignant neoplasm when patients were treated by proton therapy as opposed to photon therapy. When considering how much years of productivity is lost by a single person we should take into account Years of Potential Life Lost (YPLL) due to cancer. Li et al. (2010) show for instance that there is much room for improvement since people with an urological cancer have on average a 14.4 YPLL per death. Lundkvist et al. (2005) [2] were the only ones to include productivity losses in their cost-effectiveness study. To avoid double counting of lost productivity they only assumed IQ loss to determine productivity loss in pediatric brain tumours. Based on several studies they estimated the average IQ loss to be 17 points, 25% of this loss could be related to radiation therapy. They subsequently found a 88% risk reduction of IQ loss for patients receiving proton radiation.

3.2 Capital budgeting

Capital budgeting literature discusses six main methods to compare projects, these methods help to make investment decisions (Clark, Hindelang, and Pritchard, 1989). Four of the six are called discounted cash flow (DCF) methods because they consider the time value of money, these are discussed in paragraph 3.2.2 The two most widely used methods are however the non-DCF techniques ‘Payback period’ and ‘Return on Investment’ discussed in paragraph 3.2.1 (Lam et al., 2012). Paragraph 3.2.3 elaborates on two alternative methods used in capital budgeting, multiple valuation and real options.

3.2.1 Non-DCF methods

These techniques need little explanation, they are very simple to use, quick to calculate and the results are easy to understand. These arguments explain their wide spread usage in evaluating projects. Their simplicity often brings several disadvantages and shortcomings, therefore usage of this method in conjunction with another evaluation technique is recommended.

Payback period

(18)

17

and Pritchard, 1989). This method is useful to alert management to a long-term commitment of capital. This method however only concentrates on the recovery of capital and hence completely ignores the size of capital and its corresponding profitability. In addition revenues beyond the payback period are ignored. It also fails to consider the time value of money. Lastly, it focuses on a project’s liquidity rather than a firm’s liquidity as a whole. In my opinion there is little value in knowing the payback period of proton therapy because 1) no cost-effectiveness studies were performed this way which makes comparison impossible and 2) from a societal perspective it basically does not matter how long recovery takes since societal decisions should exceed individual time horizons taking into account future generations. Also, payback does not seem valuable in this economic evaluation because complete recovery of the initial investment is not expected.

Return on investment

The return-on-investment (ROI) method compares the investment in an asset with the annual after-tax or pre-tax income. The primary shortcoming of this approach is that the timing of expected profits is completely ignored. Thus, a project with low initial profitability and high future profitability would have the same average return as a project with higher initial profits and lower future profits. Additionally this approach uses book values rather than market values or productive asset values. This might yield extremely misleading ROI figures. In considering this economic evaluation ignorance of the timing of profits seems a big problem since timing of cash in- and outflows can substantially differ between therapies. Furthermore there is no benchmark for project acceptance like in payback period or DCF calculations. However, even if above mentioned problems can be overcome, the societal perspective taken here makes implementation of this method very difficult, if not impossible.

3.2.2 DCF methods

Discounted cash flow models are based on the concept of discounting cash inflows and outflows to their present values. Therefore these methods fully consider the time value of money. According to Wilkes and Samuels (1991) the ‘Net present value’ and ‘Internal rate of return’ are the most popular DCF methods, nevertheless this paragraph discusses the relevance of all four DCF methods.

Net present value

(19)

18

B. In specific contexts like this it might be difficult to give an accurate estimate of the appropriate discount rate, therefore an extensive discussion will follow on the cost of capital. Furthermore, in order to estimate the value of a project, complex calculations are required since every single cash transaction needs to be estimated. Inaccurate assumptions lead to inaccurate calculations of the net present value of the project. Also, NPV may not give correct decisions when projects are of unequal life. Regarding health care the NPV method seems an appropriate method since it allows for sound comparison between investment options. If necessary, adjustments for unsystematic risk, market risk, or capital structure risk can be made to the discount rate. Previous cost-effectiveness studies in health care used discounting (Sharma et al., 2001; Kauf et al., 2009). In addition, discounting has shown to be common practice in radiotherapy literature as well (Lundkvist et al., 2005 [1]; Lundkvist et al., 2005 [2]; Ramaekers et al., 2012; Quik et al., 2014). Therefore adoption of this capital budgeting method allows for comparison with previous studies.

Probability index

The probability index (PI) is the ratio of the present value of the cash inflows to the outflows. If both costs and benefits are equal the index will show the value ‘1’. If cash inflows > cash outflows (cash inflows < cash outflows) the ratio will be >1 (<1). The PI formula is shown in appendix B. The PI is a measure of a project’s profitability per dollar of investment. The main shortcoming of this technique is that it completely ignores the size of the project. This method is slightly different from the previously discussed NPV method, the only difference is that the NPV method adds cash in- and outflows while the PI divides them. In this particular setting the PI seems however not of added value since profitability is not the main concern in health care interventions. Quit often health care innovations are not profitable at all, they just boost the HRQoL of patients.

Internal rate of return

(20)

19

method inappropriate in this setting. Since we make use of incremental cash flows the cash flow sign will not be constant over time. Finally, in health care the focus is not on returns but more on cost figures and their corresponding health care effects.

Equivalent annual charge

This method involves discounting all cash in- and outflows to the present and determining the equivalent annual charge over the project’s life. This method is especially useful in evaluating nondiscretionary expenditure alternatives that are not profit producing, and for comparing projects having unequal lives (Clark, Hindeling, and Pritchard, 1989). Since we compare both therapies on a single patient basis the patient’s life can be considered as the ‘project’s life’. Because we don’t know how many years a patient will be living it seems impossible to determine the annual charge.

3.2.3 Alternative methods used in capital budgeting

Koller et al. (2005) recommend considering two alternative valuation techniques used in capital budgeting decisions: multiples (comparables) and real options. Both methods are often used in addition to one of the previously discussed methods. Where the multiples analysis provides more of a check to evaluate the accuracy of the DCF method, the real options approach really adds concepts to the valuation omitted by the DCF method.

Multiple valuation

Multiple valuation is a technique to place your DCF calculation in a proper context and is often referred to as a relative valuation method. It gives a ‘quick and dirty’ estimate of the value of a company. Since this study aims to value a medical therapy rather than a company, multiple valuation seems inappropriate. In addition, to apply a multiple valuation one needs to have comparable companies which are absent in this particular case. Even if publicly listed foreign hospitals (e.g. USA hospitals) can be found comparison may still seem impossible since radiation therapy is only a minor part of health care services provided by hospitals.

Real options

(21)

20

change course in response to changing circumstances creates real, or growth, options. Koller et al. (2005) discuss two slightly different approaches that capture flexibility: real-option valuation (ROV), based on formal option-pricing models, and decision tree analysis (DTA). Both approaches boil down to forecasting future cash flows contingent on the future states of the environment. ROV makes use of option-pricing models, the basic idea is that a portfolio of priced securities is constructed that perfectly matches the cash flows stream of the option of the project. If one succeeds, the portfolio and the option should have the same value. DTA involves discounting the contingent future cash flows at the project’s cost of capital. In principle this leads to the right answer, but only if the appropriate cost of capital for the contingent cash flows of the project is determined (Koller et al., 2005).

When considering the specific context of this study real options however seem far less useful. If the proton facility will be built it seems very unlikely that the treatment will be abandoned or expanded in the near future. Follow-on investments are already determined and taken into account in the NPV valuation. Along the way, the extensive R&D project PARTREC will recommend minor changes to the treatment based on empirical findings. These changes are however negligible and are impossible to be estimated up front. The only real option that might have a significant effect on this valuation is the option to defer investment. Deferring the investment in proton therapy for a couple of years brings the possibility of substantially lowering capital expenditures (i.e. building and equipment costs) due to increased competition and economies of scale of suppliers. Deferring investment in medical innovation is however in conflict with the mission and vision statements of the UMCG and is therefore no ‘real option’ to the UMCG.

3.3 DCF input

(22)

21 3.3.1 Cash flows

The first challenge associated with the DCF method involves the determination of cash flows. Future performance is often measured in cash flows. These cash flows are the market’s expectations and determine current value. Investors therefore aim to select strategies that maximize the present value of future cash flows (Koller et al., 2005).

Cash flows in business

Koller et al. (2005) differentiate between five DCF methods to value a company. Each method with its own cash flow and discount rate calculations. Broadly speaking one can say that adjustments for risk can either be made in the numerator (i.e. cash flows) or in the denominator (cost of capital) of the DCF method. The value of operations equals the discounted value of future free cash flow. Where ‘free cash flow’ is the cash flow available to all investors, and is independent of leverage (Koller et al., 2005). Since the cash flows of tomorrow are not certain, projections of future financial statements should be made with revenue growth and ROIC estimates. Subsequently free cash flows can be calculated with the formulae given in appendix B.

Cash flows in health care

Cash flow calculations in health care cost-effectiveness studies are completely different from free cash flow calculations in business and therefore need special attention. Unlike cost-benefit analyses it does not aim to demonstrate a project’s economic viability by comparing costs with revenues. Cost-effectiveness studies rather compare discounted cost figures with their corresponding (non-monetary) discounted effects (Lundkvist et al., 2005 [1]; Lundkvist et al., 2005 [2]; Ramaekers et al., 2011; Quik et al., 2014). Dependent on the perspective of analysis different cash flows will occur. Adopting a societal perspective basically requires to include all cash flows to all parties that are affected by a new therapy. Other perspective may focus on just a selection of cost drivers and are therefore less comprehensive.

(23)

22

substantiated. The second driver requires thorough assessment of the costs of all five side-effects. This seems to be a weakness of previous literature as well since estimates are far from comprehensive and omit relevant cash flow drivers. Lundkvist et al. (2005) [1] adopt a societal perspective and therefore added travel and hotel costs. They add a onetime cash flow at t=0 to patients receiving proton radiation because they are expected to experience substantially higher travel and hotel costs due to remote proton facilities. Lundkvist et al. (2005) [2] also adopted a societal perspective and include productivity loss in addition to previous cash flows. Ramaekers et al. (2012) and Quik et al. (2014) adopted a health care perspective and therefore only included health care cost drivers mentioned before. Since future health care costs are expected to occur evenly throughout the year, Quik et al. (2014) were followed meaning costs are discounted on a weekly basis.

Health care effects on the other hand are non-monetary figures. Nevertheless, future HRQoL differs substantially as a result of the occurrence of complications, therefore health care effects should be discounted. Dirix et al. (2007) and many others claim that especially the occurrence of xerostomia and dysphagia significantly influence the HRQoL. Previous literature examined the experienced HRQoL, expressed in utility weights ranging from 0 (death) to 1 (full health), for each complication by gathering data with the Euroqol-5D and QLQ-C30 questionnaires (Hammerlid et al. 2001; Hammerlid and Taft, 2001; Nguyen et al., 2005; Ramaekers et al., 2011). Utility weights are then multiplied by the chance of occurrence of a certain complication in both therapies. The outcomes are discounted on a weekly basis resulting in total Quality-Adjusted Life Years (QALYs) for each therapy (Visjnic et al., 2011). The ICER represents the costs per QALY, both expressions are therefore used interchangeably.

3.3.2 Cost of capital

When cash flows associated with a project are estimated, the second major challenge of the DCF method pops up, this involves determining the discount rate or cost of capital. A discount rate reflects both the time value of money and risk and therefore represents the cost of capital. In economic terms, the cost of capital for an investment is an opportunity cost – the cost of forgoing the next best alternative investment (Pratt, 2003).

Cost of capital in business

(24)

23

risk in the capital asset pricing model (CAPM). According to capital market theory, unsystematic risk can be eliminated by holding a well-diversified portfolio of investments. Consequently, the only risk that will be rewarded with a risk premium will be the asset’s systematic or unavoidable risk. The CAPM formulae is shown in appendix B. When a company is levered the most common method to calculate the required rate of return is by using the weighted average cost of capital (WACC). The WACC is a weighted average of the cost of debt and the cost of equity (Koller et al., 2005). The formula to determine the WACC can be found in appendix B. The above mentioned calculations can also be performed for listed companies. When a certain company is not listed the beta estimate should be derived from peer companies by a four step procedure discussed in appendix B. CAPM does however not seem to be appropriate for determining the discount rate for health care costs and effects. Health care costs and effects are considered from a societal perspective, this makes it hard maybe even impossible to determine the cost of capital. Therefore the next section elaborates on the discussion around discounting in health care.

Cost of capital in health care

The term ‘cost-effectiveness’ is actually a merger of two distinct terms namely costs and effects. By comparing the two, one can decide on whether the additional effects outweigh the additional costs. In the past the National Health Care Institute proposed an equal discount rate for costs and effects of 4% (National Health Care Institute, 1999). However after extensive debates in literature, summarized in Brouwer en Rutten (2005), the Health Insurance Board updated their 1999 report in 2006 and recommended the use of different discount rates for costs (4%) and effects (1,5%) (National Health Care Institute, 2006). The main reason to split these discount rates were the empirical findings of Gravelle and Smith (2001). They state that de value of health will, in monetary terms, increase in time. Discounting costs and effects against one rate ignores this appreciation, therefore they did an attempt to correct for this by making use of income growth, time preference and marginal utility of income data. They found a 3%-6.5% range for costs and a 1%-4% range for effects. Klok et al. (2003) replicated this study with Dutch data and found a 3.1%-5.4% range for costs and a 0.5%-2.3% range for effects.

(25)

24

represents the rate at which society is willing to exchange present for future consumption. Opportunity costs represent the true cost of using a resource since they describe the benefits you could have received by taking an alternative action. Despite of the inconsistencies between health care effects and these economic concepts, health care economists argue that health care effects should be discounted; saving a life in ten years from now does not have the same value as saving a life today.

Drummon, Stoddart, and Torrance (1988) argue that, in selecting the ‘best estimate’ of r, the following criteria should be considered: consistency with economic theory, government recommendations, and current practice (to which you might wish to compare results). Discount rates for both costs and effects will be discussed along these three criteria. Firstly, the theory of time preference proposes a rate to discount future health care costs in the range from 2-5% (Brouwer and Rutten, 2005; Krahn and Gafni, 1993). Proponents of a timeless view on society propose a lower discount rate for health care effects compared to health care costs. They argue that society is more than the current generation, therefore interests of future generations should be taken into account as well (Gold et al. 1996; Brouwer et al. 2000). This view on society exists already for over eighty years, Pigou (1932) was among the first to state: ‘there is wide agreement that the state should protect the interests of the future in some degree against the effects of our irrational discounting and of our preference for ourselves over our decendants’. Secondly, the National Health Care Institute recommends the use of a 4% discount rate for costs and a 1.5% discount rate for effects in health care economic evaluations (National Health Care Institute, 2006; National Health Care Institute, 2010). Additionally they recommend to perform a sensitivity analysis with a 3-5% range in order to improve international comparability. Thirdly, recent publications used discount rates in the range of 3-5% for costs and 1.5% for effects as well (Sharma et al., 2001; Kauf et al., 2009; Quik et al., 2014; Lundkvist et al., 2005 [1]; Lundkvist et al., 2005 [2]).

3.4 Conceptual model

(26)

cost-25

effectiveness research in the medical world. If will also guide future cost-effectiveness research in this field which will enhance comparability.

(27)

26

4. Data and methodology

The core of this paper is considered in this chapter. It provides a complete and accurate description of raw data and how it was compiled and analysed. In addition, used equipment and techniques are discussed in detail below.

4.1 Data

Primary input data for the simulation model was collected from the UMCG. Patient data was needed to compare proton and photon radiation therapy. Cost data was required to estimate the cash flows belonging to each side-effect. In order to determine the cost-effectiveness, HRQoL data from literature was used.

4.1.1 Patient input

An updated version of patient data used by Quik et al. (2014) was used in this study. To verify whether physical dose advantages of proton therapy result in lower incidence of side-effects, in silico planning comparative studies were performed at the UMCG for 50 patients. Only patients with a swallow sparing IMRT treatment were included. These patients are selected from an observational patient cohort of 1,008 patients treated with photon therapy, patients were excluded based on the following criteria:

- Invasive surgery before irradiation. - Special radiation scheme.

- If too much metal was present in the mouth, current techniques do not give reliable dose distribution results for proton plans.

(28)

27

As mentioned in section 2.1, proton therapy is insured care for Dutch citizens with standard or model-based indications. Only these patients are eligible because capacity is scarce and these indications are expected to reduce the risk of side-effects most (i.e. highest ∆NTCP). Therefore it is only relevant to determine the cost-effectiveness of proton therapy for patients with standard or model-based indications. The Dutch national platform proton therapy released a consensus document on 15 December 2014 in which they specify norms to determine model-based indications (Landelijk Platform Protonen Therapie, 2014). These norms are:

- Mild complications are not considered model-based.

- Moderate complications become model-based if ∆NTCP is at least 10%. - Serious complications become model-based if ∆NTCP is at least 5%.

- Life threatening complications become model-based if ∆NTCP is at least 2%.

Assisted by expert opinion it was decided that patients should at least have a ∆NTCP of 10% for one of the five complications to be eligible for proton therapy. This decision however resulted in the selection of 70% of the patients. Therefore I chose to restrict the patient cohort even more by only selecting patients with ∆NTCP of at least 20% for one of the five complications. Two reasons underlie this decision, 1) this study aims to compare two therapies, it is therefore beyond the scope of this study to extend the literature with a dataset of empirical findings on individual patient cost-effectiveness 2) in reality, only patients expected to benefit most from proton therapy are treated, therefore it is of no use to determine the cost-effectiveness of 70% of people because capacity is far from sufficient. This decision resulted in the selection of nine patients, patient characteristics of the cohort (n=50) as well as the selected patients (n=9) are shown in table 2.

Insert table 2 about here

4.1.2 Cost input

(29)

28

followed in their procedure to calculate complication costs. In short they multiplied the following three figures: % of patients using the care unit, average volume of those who use it, and cost price. The % of patients using the care unit and average volume of those who use it are based on expert opinion. Expert opinion was always verified with a second opinion. Most cost price figures, except dental care costs, are based on billable tariffs to insurers. Medical interventions or treatments are represented by DOT tariffs, medication costs are estimated by billable prices of pharmaceutical companies. Dental care costs are either based on expert opinion or on previous internal research done because no DOT tariffs are available. Prices are in Euro (€), price indices from the central statistics office (www.cbs.nl)

are used to convert all prices to 2013 levels. The simulation model tracks each patient’s complication costs, applying a 4% annual discount rate yields total costs over the prespecified time horizon.

4.1.3 HRQoL input

The extensive literature review showed consensus in cost-effectiveness studies to measure the HRQoL in terms of QALYs. This requires us to gather information about the HRQoL for each complication. By combining previous literature with expert opinion, Quik et al. (2014) assigned HRQoL weights (0= dead and 1= full health) to all complications. The maximum HRQoL of a patient having experienced cancer is assumed to be 0.9. QALYs were computed by multiplying the HRQoL weights by their duration. The simulation model tracks each patient’s QALYs, applying a 1.5% annual discount rate yields total QALYs over the prespecified time horizon.

4.2 Methodology

(30)

29

4.2.1 Economic evaluation

As shortly touched upon before, a cost-effectiveness study is, in contrast to cost-benefit or cost-utility studies, most appropriate in this particular health care setting. The benefits of a new medical therapy are hard, sometimes even impossible, to quantify in terms of money. Therefore they are measured in terms of health care effects. In this way the ICER was computed by dividing the incremental costs by the incremental QALYs. The ICER represents the costs of an additional QALY gained when comparing two strategies, it therefore provides a means of comparing projects or interventions across various medical disciplines. Whether one considers a treatment strategy to be cost-effective depends on how much society is willing to pay for every QALY gained. Ramaekers et al. (2012) were followed in adopting the informal ceiling ratio of 80,000 per QALY.

4.2.2 Simulation model Structure and assumptions

The study of Quik et al. (2014) was mainly interested in whether protons could prevent side-effects. Since first level evidence is lacking they constructed a simulation model which forms the basis of the current study. Main assumption of this model is that disease progression and thus survival rates are equal for both therapies. This assumption is based on equal tumour doses in ISPC studies. The model consists of a module for primary tumour prognosis and separate parallel modules for all relevant side-effects. For the current study, assumptions underlying the first module were unchanged while substantial alterations were made to the modules for side-effect, regarding cost estimates. All modules run continuously in time and track patient outcomes at weekly intervals with a ten year time horizon. Besides altered modules, a larger cohort of patients were simulated and cost input data was revised and updated.

(31)

30

characteristics and on time since diagnosis. For complete reasoning underlying the transition probabilities we refer to Quik et al. (2014).

Insert figure 3 about here

All five relevant complications after radiotherapy in HNC (i.e. sticky saliva, xerostomia, dysphagia, tube feeding, and hypothyroidism) were included separately and can either occur in isolation or simultaneously (figure 4). Published NTCP models were used to compute initial incidence of complication states. Details on these NTCP models are given in appendix A. A literature search performed by Quik et al. (2014) assessed long-term prognosis of complications, they found contrasting findings on late incidence and remission probabilities. These probabilities are therefore assumed to be zero.

Insert figure 4 about here

Individualized simulations

For selected patients, as discussed in the ‘patient input’ section, independent simulations were performed to predict future life courses using the model of Quik et al. (2014). Table 3 displays the model input parameters for each patient. Each simulation can be seen as a set of 1,000 identical patients simulated 1,000 times with photon and 1,000 times with proton therapy. This will give a sufficient number of result and therefore stable outcomes with low random variability. AnyLogic

®

software was used by Quik et al. (2014) to design this model.

Insert table 3 about here

Sensitivity analyses

(32)

31

5. Results

The main aim of this study was to evaluate the cost-effectiveness of proton radiation therapy, this was done by using UMCG as case study object. In addition this paper wanted to guide future research in this field by constructing a conceptual model that gives meaning to all concepts and objects involved in cost-effectiveness research of medical interventions. In order to do this properly, chapter two defined the research problem and phrased the corresponding research question and sub-questions. The sub-questions were constructed in a deductive manner, meaning answering all the sub-questions will enable me to answer the main research question. The literature review in chapter three answered the sub-questions already to a great extent. While this chapter is especially focused on answering the main research question, the answers to the sub-questions will be shortly summarized here as well.

(33)

sub-32

question was devoted to the input variable (i.e. cash flows and discount rates) of the NPV method. Cash flows are represented by, both initial and ongoing, costs from side-effects. The discount rate for indirect health care costs (4%) is different from the rate for health care effects (1.5%). The literature review elaborates on the decision to take these rates. The rest of this chapter is devoted to answering the main research question: Is proton radiation therapy cost-effective when compared to currently used photon radiation therapy?

5.1 Model input

After having found answers to all sub-questions the required input figures were collected to answer the main research question. Properly inserting these figures in the adjusted model of Quik et al. (2014) yielded outcomes on the cost-effectiveness of proton radiation therapy.

5.1.2 Cost input

Direct health care costs – treatment tariff photons

The DOT-tariff in use in the UMCG is a break-even tariff since it only recovers costs. A DOT-tariff is a collection of a least one and often several care operations. Total hospital costs (capital and operational expenditures, utilities et cetera), as stated in the annual report, are distributed to all care operations. The DOT-tariff is calculated by dividing total costs of all operations in a DOT by the number of treatments. In this way the billable DOT-tariff to insurers include all costs involved in operating a hospital and hence can be considered as a break-even tariff. The DOT-tariff for IMRT (photon radiation) is €14,446. This top down approach of cost allocation should however be interpreted with caution because cost allocation methods differ substantially between hospitals. In spite of its weaknesses, this tariff is very suitable to use in this study because this study is taken from a health care insurers perspective.

Direct health care cost – treatment tariff protons

(34)

33

provide state of the art health care services, this tariff is a break-even tariff. Below I will discuss the tariff construction process by highlighting the most relevant decisions and assumptions.

The tariff is calculated is such a way that all costs are recovered and all debt providers earn a reasonable return on investments. Proton therapy treatment time differs between eligible tumour types, therefore the business case uses an average treatment time per patient based on the case mix of indications eligible for proton therapy. Required time per patient is based on a treatment time of 23.7 radiation fractions of 24.5 minutes. Assumed is that the PTCG has a treatment capacity of 600 patients a year. A ramp-up period of three years is taken into account meaning that the facility is fully operational from the fourth year onwards. The most up to date treatment tariff for a proton radiation treatment in the case study institution is €XXXXX1. Table 4 gives an overview of the composition of the tariff.

Insert table 4 about here

The staff costs include labour costs of medical experts like anesthesiologists, radiologists but also supporting medical and administrative labour costs. Patient related material costs include material costs and diagnostic costs. Maintenance, updates & upgrades include costs of service agreements for maintenance to devices, equipment, and software as well as costs for annual updating and upgrading of equipment. Other operational costs include costs for utilities, insurances, and facilitating management. Lastly, capital expenditures are composed of initial and reinvestments. These investments comprise of costs for building, technique, cyclotron, gantries and other investments. Profit and loss accounts, cash flow statements and balance sheet projects are made for the full operating period of 30 years in order to determine financing needs. Depending on the available cash flows debt holders receive repayment and compensation for their funds.

Indirect health care-costs – complication costs

Hakkaart, Tan, and Bouwmans (2010) wrote a guide for research into the costs of health care. Their procedure to calculate health care costs was followed here. Table 5 gives an overview of the health care costs to insurers of the five complications. In identifying complication costs I differentiated between one-off costs from medical intervention and ongoing (yearly) costs from supportive care,

(35)

34

medication, and repeating treatments. The yearly health care costs to insurers varies considerably between complications from about €30 to €11,000.

Insert table 5 about here

Except for some simple medication not much can be done to cure sticky saliva. About half of the patients experiencing xerostomia is expected to take artificial saliva (Saliva Orthana, Glandosane, and Xialine). An indirect link is made to dental care costs, from xerostomia patients 20% is expected to receive implants before radiation therapy because the risk to damage teeth and gum is too high. Also, previous internal research shows that 33% of xerostomia patients receive a partial or complete denture after radiation therapy. Xerostomia patients having received either implants or a denture are all expected to see the jaw surgeon once a year and the hygienist four times a year. Only a very small part (5%) of patients experiencing dysphagia is expected to use liquid food (Nutricia nutridrink protein or Cenaman protein energy) because pureeing your food is often sufficient. Approximately 15% of people needs oesophagus dilatation as a result of scarring. All dysphagia patients are expected to see the dietician once a year to monitor eating habits. Tube feeding, the most severe state of dysphagia, gives no other choice then to eat through a tube. The majority (80%) receives a tube directly to the stomach which is implanted in a surgical intervention. The other 20% gets a nose tube for which one-off costs are negligible and therefore ignored. Patients experiencing hypofunction of the thyroid receive medication (Levothyroxine), see a general practitioner now and then, and are monitored with lab tests to see whether the functioning of the thyroid changes. Only a small part (10%) is expected to see an endocrinologist once a year.

5.1.2 HRQoL input

(36)

35

Insert table 6 about here

Based on expert opinion it was decided that the HRQoL of sticky saliva is equal to the HRQoL of xerostomia because the both involve saliva problems. Hypothyroidism is expected to have a minor influence on the HRQoL while tube feeding dramatically reduces the HRQoL.

5.1.3 NPV model

The NPV model underlying the simulation is essential in that it allows us to compare both therapies at one point in time by discounting future health care costs. The NPV model can best be explained by a simplified example as shown in appendix C. Let’s consider the first five years of hypothetical patient’s life after photon radiotherapy and compare this with the situation in which this same patient was treated with proton radiotherapy. Table C1 presents input assumptions underlying the NPV calculations of health care costs shown in table C2. Subsequently table C3 displays input assumptions for the NPV calculations of health care effects shown in table C4. In addition, all NPV calculations are also presented in formula format using formula (3) from appendix B. Table C2 illustrates that discounting future health care costs reduces the cost difference between both therapies from €15,000 (€30,000-€15,000) to €9,126 (€35,874-€26,748). Besides, table C4 shows a gain in QALYs of 0.19. This hypothetical patient will therefore be eligible for proton radiation therapy since costs per QALY gained are €48,032.

5.2 Outcomes

5.2.1 Standard case

(37)

36

Detailed outcomes of one patient (number one) are discussed here, simulation outcomes of the other eight patients are summarized in table 7. A standard simulation was performed over a time horizon of 10 years, with discounts of 4% and 1.5% for respectively health care costs and health care effects, and with complication costs as shown in table 5. Having simulated 1,000 identical patients with characteristics of patient one (figure 5) shows that many of them are dead or experienced disease progression. Photon simulations showed that from the 234 (198-275) patients alive (disease-free), 102 (70-136) had sticky saliva, 148 (115-182) suffered from xerostomia, 37 (21-54) had less severe dysphagia, 73 (52-93) needed tube feeding, and 135 (106-173) were diagnosed with hypothyroidism. Proton simulations on the other hand showed significant improvements. From 233 (188-280) patients alive (disease-free), 91 (66-119) had sticky saliva, 124 (92-164) experienced xerostomia, 12 (3-25) had less severe dysphagia, 42 (25-61) needed tube feeding, and 122 (93-153) were diagnosed with hypothyroidism.

Graphs on the right-hand side depict the expected probabilities of having a complication ten years after treatment for both radiation therapies. The (non)prevalence of these complications directly influences a patient’s HRQoL as well as indirect health care costs. Therefore the complication outcomes are combined into QALYs and costs graphs which are depict on the left-hand side. The x-axis shows counts of simulation outcomes in specified intervals, the y-axis displays these counts as a percentage total simulations.

Insert figure 5 about here

We saw in section 5.1.2 that treatment costs of proton therapy are €XXXXX1 (€XXXXX2- €14,446) higher than treatment costs of photon therapy. This study shows that for patient 1 the difference between total costs (i.e. treatment and complication costs) is reduced to €10,466 (€37,244 - €26,778). Total QALYs are 3.419 and 3.265 for respectively proton and photon radiation therapy. From this a ∆QALYs of 0.154 (3.419 – 3.265) is calculated. The ICER, also referred to as costs per QALY, of patient 1 is therefore €67,840 (€10,466/0.154). Table 7 shows for each simulated patient the ∆Costs and ∆QALYs. By dividing the former by the latter we find ICER figures.

1 ∆Treatment tariffs will give away proton treatment tariff, hence it was only announced to supervisors

(38)

37

Insert table 7 about here

When reflecting these outcomes in light of the informal ceiling ratio of 80,000 per QALY adopted by Ramaekers et al. (2012), we see that proton radiation is cost-effective in one third (i.e. patients one, two, and six) of the simulated patients. Another third (i.e. patients three, five, and nine) of the patients balances on the edge of being regarded as cost-effective while it is not cost-effective to simulate the remaining three patients (i.e. patients four, seven, and eight) with proton therapy. The expectation that proton therapy is seen as a cost-effective treatment in three out of 50 patients is a very promising result. This result is especially promising when kept in mind that a lot of conservative assumptions are made regarding proton therapy. The disappointing results of three patients for whom proton therapy is far from cost-effective can be explained by the selection criteria rather than the therapy itself. This study shows that proton therapy is cost-effective only if appropriate selection criteria are used. The results are discussed more extensively in the next chapter.

5.2.2 Sensitivity analyses

(39)

38

6. Discussion

(40)

39

The majority of studies comparing photon and proton radiation therapy were ISPC studies only reporting dose and volume reductions. The power of a simulation model originates from combining ISPC study outcomes with NTCP models to estimate the reduced probability of a certain complication. Four studies performed a cost-effectiveness evaluation of proton therapy for HNC (Lundkvist et al., 2005 [2]; Ramaekers et al., 2012; Quik et al., 2014). Lundkvist et al. (2005) [2] reported extremely low costs per QALY gained of €4,000. Thereafter, Ramaekers et al. (2012) found total costs of €60,000 for each QALY gained. Quik et al. (2014) mainly focused on the development of a simulation model to provide expected outcomes for individual patients. While they did not report the costs per QALY gained, they reported results from which costs per QALY of €148,000 can be calculated. This study presents costs per QALY between €60,000 and €70,000 for three patients for whom proton therapy is considered cost-effective. The remarkable result of Lundkvist et al. (2005) [2] can be explained by their assumption that proton therapy is able to improve tumour control and prevent complications at the same time. As argued before, the present study follows Ramaekers et al. (2012) and Quik et al. (2014) by assuming either one of the advantages can be exploited depending on the treatment plan. Ramaekers et al. (2012) took a lifelong time horizon while Quik et al. (2014) modeled only the first five years after treatment. The time horizon of 10 years used here might explain why the costs per QALY are between dose found by Ramaekers et al. (2012) and Quik et al. (2014). All three studies exploring the cost-effectiveness used a ‘typical’ or ‘average’ patient, in contrast nine patients are simulated and analyzed here.

The perspective of analysis strongly influences the input and therefore the outcomes of cost-effectiveness studies. Interpreting the results in light of outcomes of others should therefore be done with caution. Where Lundkvist et al. (2005) [2] used a societal perspective, Ramaekers et al. (2012) and Quik et al. (2014) adopted a health care perspective. This general health care perspective aims to estimate the real costs associated with complications. The present study however approached the health care perspective from a different angle by focusing on relevant health care costs to insurers. This was done because health care insurers are profit-seeking firms and real cost estimates are therefore meaningless to them. While these different perspective require different input parameters (i.e. discount rates and cost estimates), the underlying model, as shown in figure 2, remains the same.

(41)

40

be used in practice. Secondly, disease progression and long-term prognosis of complications are uncertain factors that influence the outcomes. Third, direct as well as indirect non-health care costs are ignored which reveals a gap in the evidence. Fourth, a time horizon of only 10 years was considered, it would however be optimal to track patients during their entire life after radiation therapy. Fifth, patients experiencing events of their primary tumour were excluded from further tracking since additional treatment would interfere with complication risks. Lastly, comparison between photon and proton therapy can be seen as unfair since the former already went through several decades of development, improvements and cost reductions while the latter did not. One has to keep in mind that innovations are expensive but have a much greater scope for improvement in the near future. The preceding four limitations can be regarded as conservative towards proton therapy.

Referenties

GERELATEERDE DOCUMENTEN

• Unbalance in loading, asymmetry in supply voltages, AND distortion in voltage and/or current contributes to the degradation of power factor (the effiency in the.. transfer of

Uit bovenstaande blijkt dat er geen significant bewijs is gevonden voor de eerste hypothese dat tight budget control een negatief effect heeft exploratieve innovatie.. De

The study has tried to answer the question: How can the relevance and quality of curricula, examinations, and assessment be improved at the classroom level in junior and senior

Management control theories are used to elaborate the concept of management control and to describe management control practice at Unilever. Furthermore cash flow

From the standpoint of cost-effectiveness, centralizing the nine community hospitals to two stroke centers seems to be the economically most attractive organizational model

Consistent with neoclassical realist expectations, the willingness to reduce the domestic political costs of deploying military personnel abroad has played a key role in

More specifically, we explored how the perceived quality of online relationships with work-related Facebook contacts and the perceived authority or power of such contacts is