Eur J Health Econ DOI 10.1007/s10198-016-0763-8
ORIGINAL PAPER
Preferences for antiviral therapy of chronic hepatitis C: a discrete choice experiment Axel C. Mu¨hlbacher1,2 • John F. P. Bridges3 • Susanne Bethge1 • Ch.-Markos Dintsios5,6 • Anja Schwalm4 • Andreas Gerber-Grote4 • Matthias Nu¨bling7
Received: 19 March 2015 / Accepted: 11 January 2016 Ó Springer-Verlag Berlin Heidelberg 2016
Abstract Background The German Institute for Quality and Efficiency in Health Care (IQWiG) uses patient-relevant outcomes to inform decision-makers. Objective IQWiG conducted a pilot study to examine whether discrete choice experiments (DCEs) can be applied in health economic evaluations in Germany to identify, weight, and prioritize multiple patient-relevant outcomes, using the example of antiviral therapy for chronic hepatitis C (HCV). A further objective was to contribute to a more structured approach towards eliciting and comparing preferences across key stakeholders. Methods In autumn 2010, a DCE questionnaire was sent to patients with chronic HCV to estimate preferences across seven outcomes (‘‘attributes’’), including treatment efficacy [sustained viral response (SVR) at 6 months],
& Axel C. Mu¨hlbacher [email protected]
adverse effects (flu-like symptoms, gastrointestinal symptoms, psychiatric symptoms, and skin symptoms/alopecia), and measures of treatment burden (duration of therapy, frequency of injections). A linear model and an effects coded full model were applied to assess the relative importance of the attributes. Results In total N = 326 patients were included. A clear preference for SVR was shown; frequency of injections and duration of therapy shared the second rank, while psychiatric symptoms ranked third. The duration of flu-like symptoms was the least important attribute. Conclusion Our findings indicate that it is possible to perform a DCE at the national level in a health technology assessment agency. The weighting of multiple outcomes allows an indication-specific and evidence-based measure to be used in health economic evaluations. In decisionmaking in health care, the approach generally allows for consideration of patient-relevant trade-offs regarding the benefits and harms of medical interventions. Keywords Discrete choice experiment (DCE) Conjoint analysis (CA) Hepatitis C virus (HCV) Health technology assessment (HTA) Patient preferences Priority setting
1
Institute for Health Economics and Health Care Management (IGM), University of Applied Sciences Neubrandenburg, Brodaer Straße 2, 17033 Neubrandenburg, Germany
2
Center for Health Policy and Inequalities Research, Duke Global Health Institute, Duke University, Durham, NC, USA
3
Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA
JEL Classification
4
Institute for Quality and Efficiency in Health Care (IQWiG), Cologne, Germany
Introduction
5
Strategic Market Access Intelligence, Bayer Health Care, Leverkusen, Germany
6
Department of Public Health, Faculty of Medicine, Heinrich-Heine University, Du¨sseldorf, Germany
7
Empirical Consulting mbH (GEB mbH), Freiburg, Germany
Health policy decision-makers and clinicians largely draw their conclusions on the benefits and harms of medical interventions on the basis of different outcomes investigated in clinical trials. However, due to the large number and diversity of these outcomes, it is a highly complex task
I 11 H 51 I 18
123
A. C. Mu¨hlbacher et al.
to determine the true ‘‘value’’ of competing interventions. Even if specific decision criteria are established, their relative importance to patients and other stakeholders remains unclear. The use of multiple decision criteria therefore calls for the weighing of benefits and harms to inform decisionmaking [1]. A health technology assessment (HTA) using multiple outcomes from a clinical trial can thus only provide comprehensive information for decision-making if these outcomes are weighted and can be reduced to a common denominator. Various methods of multi-criteria decision-making have been tested in the past years, such as the analytical hierarchy process (AHP) and a stated preference method called, ‘‘conjoint analysis’’, including its choice-based form, the discrete choice experiment (DCE) [2]. However, no established scientific standard for the identification, weighting and prioritization of outcomes currently exists. It is therefore unclear how multiple outcomes should be handled in the decision-making process. To prepare guidance for transparent prioritization and allocation decisions, it is desirable to determine the actual value judgments of key stakeholders. In recent years, many HTA agencies have aimed to broaden their scope of evaluation by including a wide range of perspectives, including the patient perspective [3]. This approach is considered to be an important element in transparent priority setting [4]. However, up to now there has been no practical guidance on how to involve key stakeholders in HTA. The Institute for Quality and Efficiency in Health Care (IQWiG), the main German HTA agency, assesses the benefits of interventions for patients on the basis of patientrelevant outcomes. According to the German Social Code Book V (SGB V), the outcomes primarily considered are mortality, morbidity and health-related quality of life [5]. IQWiG must ensure that the assessment of the patientrelevant benefit of a medical intervention is conducted according to internationally accepted standards of evidence-based medicine and health economics [§139a (4) SGB V]. In Germany, the health economic evaluation of an intervention is supposed to be based directly on the patientrelevant outcomes of clinical trials; therefore, various methods for weighting outcomes had to be tested. In addition, IQWiG and other organizations have been promoting the position that elicitation of patient preferences is one of the most important challenges in the construction of a utility metric used in a cost-benefit framework. IQWiG therefore initiated a pilot study to examine whether a DCE can be applied in health economic evaluations in Germany to identify, weight and prioritize multiple patient-relevant outcomes, using the example of antiviral therapy for chronic hepatitis C (HCV).1 This study focused on the valuation of different decision criteria or patient relevant endpoints, but did not assess quality of life
123
or health episodes. The results of the study were also to be used to inform regulatory and health policy decisionmakers about the feasibility of the approach in the context of HTA and about the design of surveys on patient preferences. A further objective was to contribute to a more structured approach towards eliciting and comparing preferences across key stakeholders. The present pilot study was conducted in the form of a survey and in parallel to an AHP study in patients with depression [6]. The full German-language report and an English-language summary are available on the IQWiG website [7].
Methods Choice of approach In contrast to AHP methods [6], DCEs can be used to compute relative importance weights for various treatment attributes that influence the benefit for patients. DCEs are grounded in psychology [8], economics [9], and econometrics [10]. They are based on respondents’ assessments of competing alternatives within choice sets. Over the last decade, DCEs have been used to analyze respondents’ perspectives on resource allocation decisions [11–13], as well as to determine the preferences of patients, insured persons or experts in the healthcare system [14–19]. Medical interventions can be described by means of characteristics or outcomes of a treatment, which are referred to as attributes, and by the levels of these attributes. In a DCE, utility is measured on an arbitrary scale, and the overall importance of each attribute depends on the range of attribute levels chosen for the experiment. Several guidelines have been developed for the application of DCEs in health care [14, 20, 21]. Context of disease The World Health Organization (WHO) estimates that the global incidence of HCV is 3 % [22]. The disease leads to increased mortality and morbidity rates, as well as to a significant impairment in quality of life [23–28]. In 80 % of the cases, the course of disease is chronic [29]. When the pilot study was performed in 2010, the clinical guidelines 1 Note The present pilot study on the treatment of HCV uses attributes and outcomes that might not necessarily correspond to patient-relevant outcomes pursuant to SGB V (i.e., outcomes describing morbidity, mortality, and health-related quality of life). It cannot be concluded from the language used in the present paper that IQWiG would regard the attributes and outcomes applied to represent (patient-relevant) outcomes in the event of a benefit assessment.
Preferences for antiviral therapy of chronic hepatitis C
for the treatment of chronic HCV recommended a combination therapy with peg-interferon and ribavirin for 48 or 72 weeks; telaprevir and boceprevir, as well as interferonfree regimens, were not available at that time [30]. The clinical objective of antiviral HCV therapy is to achieve sustained virological response (SVR), i.e., permanent eradication of the virus. Treatment decisions in patients with HCV should generally be made taking patient preferences into account [31–38]. Besides SVR, further patientrelevant evaluation criteria should therefore be considered. Patient recruitment and selection The survey was conducted in September and October 2010 with paper-and-pencil questionnaires. Questionnaires were sent to patients with HCV who were members of the ‘‘Deutsche Leberhilfe e.V.’’ (German Liver Support Group), a large patient organization. Alternatively, they could fill out an online questionnaire provided by the same organization. Patients with HCV who were older than 18 years of age were eligible for inclusion. Patients with active alcohol or drug abuse or co-infection with human immunodeficiency virus were excluded because of possible interferences with medication. To achieve a sufficient sample size, no HCV genotype was excluded. Ethical approval All respondents were informed in writing about the study objective, as well as the potential benefits and risks, prior to participation and provided informed consent. Participation was voluntary and not reimbursed. The study, together with all information materials and the questionnaire, was ethically approved by the Medical Association of Northern Rhine. Survey components The questionnaire was structured as follows: Section 1: patient characteristics The first section contained questions on general sociodemographics (e.g., gender, age, marital status, and educational level), as well as medical history (e.g., current health status, year of first diagnosis, current treatment status, and genotype). Section 2: attributes and levels The second section contained an introduction to the attributes and levels used in the survey. To ensure that this selection was based on the best possible evidence, a
targeted literature search was conducted in March and April 2010 in the PUBMED, MEDLINE, SPRINGER, and THIEME databases to identify the most important attributes and levels. The search covered the relevance and effects of chronic HCV, as well as treatment options. The main focus of the literature search was on randomized controlled trials (RCTs) and evidence-based guidelines. In addition, two focus groups (n = 9) were formed to ensure that no relevant outcomes had been overlooked, and to examine the possible level ranges, as well as the comprehensibility of the attributes. At a later stage, the focus groups were used for the qualitative validation of the selected attributes and their values (levels) to ensure a comprehensive account of the specific assumptions of the affected patients. On the basis of the results of the literature search and the focus groups, seven attributes were extracted and included in the questionnaire. The above nine subjects pre-tested a draft version of the questionnaire, which was found to be manageable. The final attributes and levels included in the questionnaire and the underlying literature are shown in Table 1. Section 3: choice tasks The third section contained the choice tasks (see Fig. 1). Patient preferences were measured by asking which of two hypothetical treatment options respondents would choose. Within the choice tasks, they were not required to indicate the strength of their preferences or to assess the confidence they had in their choice. In order to clarify the meaning of the attributes and to ensure that these attributes were understood in the same way, an easy-to-understand, non-technical explanation was provided for each attribute. The full profile approach was applied for the construction of the choice tasks, which means that each choice set included all seven attributes. In DCE research, four to eight attributes per choice set are seen as appropriate [39, 40]. As clinical guidelines strongly recommend treatment of HCV, no ‘‘opt-out’’ or ‘‘status-quo’’ alternative was included in the choice tasks. After the choice tasks, the respondents could make their own comments and were asked to indicate how difficult it was to fill in the questionnaire. Experimental design Given that the study did not aim to explore any specific interactions, a main-effects experimental design was used. Specifically, we used an orthogonal experimental design that accounted for three attribute levels by applying a 6^7 orthogonal array, producing a D-optimal experiment [41].
123
A. C. Mu¨hlbacher et al. Table 1 Attributes and levels included in the questionnaire
Attribute Duration of antiviral therapy
Level
References
12 weeks
[29–30]
24 weeks 48 weeks Frequency of interferon injections
Three times per week
[29–30]
Once per week Once every 14 days Duration of flu-like symptoms after injection
1 day after injection 2 days after injection 3 days after injection
Probability of gastro-intestinal symptoms
25 of 100 persons (25 %) 35 of 100 persons (35 %) 45 of 100 persons (45 %)
Probability of psychiatric symptoms
35 of 100 persons (35 %) 45 of 100 persons (45 %)
Probability of skin symptoms and/or alopecia
55 of 100 persons (55 %) 35 of 100 persons (35 %) 45 of 100 persons (45 %) 55 of 100 persons (55 %)
Probability of SVR 6 months after end of therapy
45 of 100 persons (45 %) 55 of 100 persons (55 %) 65 of 100 persons (65 %)
Fig. 1 Example of a choice task
Therapy A
Therapy B
24 weeks
48 weeks
Frequency of interferon injections
once every 2 weeks
once a week
Duration of flu-like symptoms after injection
3 days after injection
one day after injection
Probability of gastrointestinal symptoms
25 out of 100 people (25%)
45 out of 100 people (45%)
Probability of psychiatric symptoms
55 out of 100 people (55%)
45 out of 100 people (45%)
Probability of skin problems or alopecia
55 out of 100 people (55%)
45 out of 100 people (45%)
Probability of SVR 6 months after treatment
55 out of 100 people (55%)
45 out of 100 people (45%)
□
□
Duration of treatment
Please choose one therapy option.
The approach (i.e., use of a higher-order design to generate both profiles in the pair simultaneously) was considered a more efficient, and potentially more accurate, approach to construct the orthogonal experimental design in the absence of priors [42]. The planned sample size according to the Orme calculation was a minimum of n = 168 participants (n = 42 per DCE Block) [43].
123
The experimental design led to 72 paired-comparison choice tasks, which were randomly assigned to four blocks of 18 choice tasks each, in order to reduce the cognitive burden for patients. The final experimental design resulted in four different versions of the questionnaire; each respondent was randomly allocated to one of these versions. All choice tasks were forced choice options, i.e.,
Preferences for antiviral therapy of chronic hepatitis C
respondents had to choose one option. They were asked to make 18 choices, as other studies have shown that this number of choices is manageable without significant differences in results [44]. The whole experimental design and all choice-set are available at: https://www.iqwig.de/down load/GA10-03_Arbeitspapier_Version_1-1_Conjoint-Ana lyse-Pilotprojekt.pdf. The testing of the experimental design confirmed 100 % D-optimality for main effects in a paired-comparison experiment, as well as orthogonality at both the profile and experiment level. Furthermore, level balance (i.e., each level was equally likely to appear on the first or second profile in the experiment) and zero overlap (occurrence of the same level in one choice task) were given. Data analysis In order to assess the relative importance of attributes, we fitted a linear model, as well as an effects coded model (full model), using SPSS 18 and STATA 11. Within the linear model, the difference between levels was coded in only one variable for each attribute. All three levels are thereby treated as linear degrees with identical distances between levels. In the full model, effects coding was used for all seven attributes. The worst level (e.g., highest probability of adverse effects) was in each case (model) coded as the reference category. Subsequently, the difference between treatment options A and B was generated and coded in two variables for the upper and the middle level. The linearity assumption of the levels underlining the linear model needs to be tested in the full model showing the estimated preference weights for all levels included. If one assumes a priori that the levels are not nominally, but ordinally scaled, it can be tested in the full model whether the linearity assumption applies across all levels of attributes. The coefficient per attribute therefore indicates the increase per level (difference between A and B). Linearity must be tested, because identical distances between levels cannot be presumed. If linearity is confirmed, the linear model suffices for the identification and weighting of attributes, as this model is adequate to describe linear relationships. In our study, we calculated to what extent level 1 takes the middle ranking in preference between levels -1 and 1. For the effects coding used, this means that the estimated coefficient of the middle level should be close to zero. Quality of the statistical model The quality of the various models was assessed and compared using the Akaike Information Criterion and Bayesian Information Criterion, which are both used to assess the general measurement quality of models by means of the likelihood ratio test. Model quality can also be assessed by
means of the proportion of choices correctly classified by the model (proportion of choice tasks with a predicted probability of choice of [50 % in the model compared with the actual choices made by patients). The full model and the linear model correctly predicted 74.24 and 74.11 % of the choices, respectively. The linear model could therefore be applied. Data interpretation Within the estimated models, the coefficients presented indicate the relative importance of an attribute, e.g., the level within the treatment choice. The higher the value of a coefficient, the higher the importance of the related attribute/level within the decision. A positive coefficient indicates that the attribute is valuable to the patient. A negative coefficient indicates that the participant attaches a negative value. The final ranking is based on the coefficient for each patient-relevant outcome out of the coefficients in the linear model. The coefficient can be interpreted as the relative value of an improvement of one level for each attribute. The full model was effect coded, and therefore displays coefficients for each single level per attribute (Table 3). We generated the mean relative importance of each attribute by computing the differences between the highest and lowest coefficients for the levels of each attribute. We then normalized the scale by assigning ten units to the most important attribute and measured the other attributes’ importance relative to this change. The mean relative importance score for each attribute expresses an improvement for an attribute on a scale from 0 (worst level) to 10 (best level); Fig. 2.
Results Respondent characteristics A total of 326 patients, 64 % of them female, participated in the survey (234 completed the paper and 92 the online version. Controlling for the mode of administration (paper or online version), no significant differences were detected.) Their mean age was 58.9 years. Within the study sample 39.9 % had already completed treatment, 7.7 % were currently undergoing treatment, 24 % had discontinued, and 28.2 % had not yet started a HCV treatment. 69.9 % were genotype 1. 60.5 % stated that it was easy or very easy to fill in the survey. The demographic and clinical data are shown in Table 2. Of the 326 questionnaires returned, 17 contained no answers to the choice tasks and could not be analyzed. In all, 5252 decisions from 309 respondents were used for the final analysis.
123
A. C. Mu¨hlbacher et al. Fig. 2 Graphic display of level estimates and normalization of coefficients
Results of the preference elicitation models The analysis of the linear model resulted in the following weights for the seven attributes: ‘‘SVR 6 months after end of therapy’’ (coef.: 0.8041), ‘‘frequency of interferon injections’’ (coef.: 0.2966), ‘‘duration of antiviral therapy’’ (coef.: 0.2502), ‘‘probability of psychiatric symptoms’’ (coef.: 0.1857), ‘‘probability of gastrointestinal symptoms’’ (coef.: 0.1233), ‘‘probability of skin problems or alopecia’’ (coef.: 0.1054), and ‘‘duration of flu-like symptoms after injection’’ (coef.: 0.1052); Table 3.
123
On the basis of the normalized difference between the coefficients of the best and worst level of the full model, the attributes ranked as follows: SVR (10), frequency of injections (3.69), duration of therapy (3.18), psychiatric symptoms (2.29), skin symptoms/ alopecia (1.48), gastrointestinal symptoms (1.44), and flu-like symptoms (1.27); see Fig. 2. If confidence intervals were taken into account, SVR ranked first, with frequency of injections and duration of therapy sharing the second rank, while psychiatric symptoms ranked third.
Preferences for antiviral therapy of chronic hepatitis C
Results of the significance tests
Table 2 Patient characteristics Item
N
N in %
Overall
326
100
Gender Female
208
63.8
Male
117
35.9
Not reported
1
0.31
Age Median (years) Not reported
58
–
7
2.15
245
75.15
20
6.13
Marital status Married/living with a partner Single Separated
5
1.53
Widowed
16
4.91
Divorced
38
11.66
2
0.61
Not reported
In the comparison of attributes, the preference for SVR was significantly higher than for all other attributes. The difference between the coefficients of ‘‘duration of therapy’’ and ‘‘frequency of injections’’ was not significant, but was significant between these two attributes and all other ones. These two attributes were therefore ranked in second place. The attribute ‘‘psychiatric symptoms’’ ranked in fourth place. ‘‘Gastrointestinal symptoms’’ (fifth place) and ‘‘skin symptoms/alopecia’’ were weighted higher than ‘‘flu-like symptoms.’’ As no significant differences were shown between these three attributes (overlapping confidential intervals), they therefore shared a conjoint fifth to seventh place.
Discussion
Net-income/pension in last month Less than 1000 Euro
108
33.13
1000–2000 Euro
124
38.04
2000–3000 Euro
49
15.03
3000–4000 Euro
12
3.68
More than 4000 Euro
15
4.60
Not reported
18
5.52
No fibrosis
67
20.55
Minor fibrosis
80
24.54
Moderate fibrosis
69
21.17
Current status of disease
Comparison with previous research
Severe fibrosis
12
3.68
Cirrhosis
50
15.34
Not known
40
12.27
8
2.45
Type I
228
69.94
Type II
30
9.20
Type III
27
8.28
Type IV
9
2.76
Type V
3
0.92
22
6.75
7
2.15
Not reported HCV genotype
Not known Not reported Diagnosis of hepatocellular carcinoma Yes
3
0.92
293
89.88
29
8.90
1
0.31
No treatment started
92
28.22
Currently being treated
25
7.67
130
39.88
78
23.93
1
0.31
No Not known Not reported Current therapy status
Treatment completed Treatment discontinued Not reported
In a sample of more than 300 patients, patient preferences were determined in a DCE for attributes of antiviral HCV therapy. Our study shows that patients are able to consider alternative scenarios for treatments and outcomes when choosing their optimum treatment strategy.
The results of the present study are in line with similar analyses in other healthcare systems. For example, a preference study by Fraenkel et al. [37] conducted in the US also found that achieving SVR 6 months after the end of therapy was the most important attribute of anti-HCV therapy. Moreover, the study showed that the decision to treat depends on the severity of potential treatment-related adverse effects. If adverse effects were expected to be mild, 67 % of patients reported that they would opt for treatment. However, if adverse effects were expected to be severe, this rate decreased to 51 %. Two further US studies (Hauber et al. [45] and Kauf et al. [46]) also found SVR to be the most important attribute for patients within a treatment decision. As the shared second rank shows, the attributes ‘‘frequency of injections’’ and ‘‘duration of therapy’’ are essential outcomes in the assessment of the benefits of antiviral HCV therapy by patients. Treatment regimens usually last between 48 and 72 weeks, and daily interferon injections seem to represent a great burden and should be taken into account in a benefit assessment [30, 47, 48]. The same applies to all side effect attributes, which should also be taken into consideration. As all attributes resulted in significant coefficients, all are relevant for patients and need to be addressed within value judgments.
123
A. C. Mu¨hlbacher et al. Table 3 Parameters estimated from the linear and full model Attributes and levels
Duration of antiviral therapy
Linear model
Full model
Coefficients
Odds ratio
SE coeff
95 % confidence intervals
0.2502
1.2843
0.0234
0.204
p
Coefficients
Odds ratio
SE coeff
95 % confidence intervals
p
–0.356 –0.248 \0.001
0.296 \0.001
48 weeks
–0.3019
0.7394
0.0270
24 weeks
0.0869
1.0907
0.0274
0.030
0.140 \0.01
12 weeks
0.2151
1.2399
0.0275
0.160
0.270 \0.001
–0.3748 0.1490
0.6874 1.1607
0.0275 0.0276
0.2258
1.2533
0.0273
Frequency of interferon injections
0.2966
1.3453
0.0234
0.251
0.342 \0.001
Three times per week Once per week Once per 14 days Duration of flu-like symptoms after injection
0.1052
1.1109
0.0232
0.060
–0.429 –0.321 \0.001 0.090 0.200 \0.001 0.170
0.280 \0.001
0.151 \0.001
3 days after injection
–0.0877
0.9160
0.0269
–0.140 –0.035 \0.001
2 days after injection
–0.0319
0.9686
0.0277
–0.090
1 day after injection
0.1196
1.1270
0.0273
0.070
45 of 100 persons
–0.1296
0.8785
0.0270
–0.183 –0.077 \0.001
35 of 100 persons
0.0250
1.0253
0.0275
–0.030
25 of 100 persons
0.1046
1.1103
0.0274
0.050
Probability of gastrointestinal symptoms
Probability of psychiatric symptoms
0.1233
0.1857
1.1312
1.2041
0.0233
0.0234
0.078
0.140
0.020
n.s.
0.170 \0.001
0.169 \0.001
0.080
n.s.
0.160 \0.001
0.232 \0.001
55 of 100 persons
–0.1820
0.8336
0.0277
–0.236 –0.128 \0.001
45 of 100 persons
–0.0086
0.9914
0.0272
–0.060
35 of 100 persons
0.1906
1.2100
0.0273
0.140
55 of 100 persons
–0.1509
0.8599
0.0274
45 of 100 persons 35 of 100 persons
0.0907 0.0602
1.0949 1.0620
0.0271 0.0268
45 of 100 persons
–0.8233
0.4390
0.0292
–0.881 –0.766 \0.001
55 of 100 persons
0.0192
1.0194
0.0256
–0.031
65 of 100 persons
0.8041
2.2347
0.0296
0.750
Probability of skin symptoms and/or alopecia
Probability of SVR 6 months after therapy
0.1054
0.8041
1.1112
2.2347
0.0263
0.0261
0.060
0.753
0.040
n.s.
0.240 \0.001
0.151 \0.001
–0.205 –0.097 \0.01 0.040 0.010
0.140 \0.01 0.110 \0.05
0.855 \0.001
0.069
n.s.
0.860 \0.001
# obs.: 5252; N = 326; LR chi2(7) = 1563.26; Log likelihood = –2852.7476; Prob [ chi2 = 0.0000; # obs.: 5252; N = 326
Importance of the study and research context When confronted with a multi-criteria decision problem (as is the case in the treatment of HCV), it needs to be taken into account that the value of (clinical) outcomes for the patient should be considered, not the actual clinical result, as has been common practice thus far. Measurability must thus be assumed not only for the difference in clinical
123
results, but also for the difference in values between the outcome levels. In the context of an allocation decision, it is important that the values of the outcomes are used. The benefit of a medical intervention for patients can only be assessed with the involvement of the affected patients themselves. As a one-dimensional indicator to explain choices about decisions, patient preferences represent the extent to which a therapy alternative should be
Preferences for antiviral therapy of chronic hepatitis C
favored from the patient perspective. The perceived or expected benefit (or harm) from this perspective is thereby the basis for patient preferences and consequently one of the explanatory factors for a patient’s choice of action. For this reason, on the basis of the known preferences of an individual or a population of patients, conclusions can be drawn about the relative benefit of a treatment alternative. Within the legal frameworks of various jurisdictions and decision-making bodies, results of such studies could feed into decisions on reimbursement and pricing of interventions. Limitations of the study With regard to the generalizability of the results, it should be noted that the study sample was characterized by strong variations of individual characteristics, even if not fully representative. It was aimed to include a broad range of hepatitis patients to be able to present a most informative model. The mean age of patients was nearly 60 years and nearly two-thirds were female; however, in Germany patients with hepatitis C are generally younger and predominantly male [49, 50]. This lack of representativeness may, among other things, be explained by the fact that patient recruitment was conducted solely via one patient organization, meaning that not all patients eligible (e.g., those in prison or other high-risk patients) could be reached. Other preference studies have shown that values of patients may depend on their cultural background and the healthcare system they live in [46]. This should also be taken into account when interpreting the study results, as the study was conducted in Germany. While interpreting the data, the linearity over the range of levels evaluated needs to be verified. If nonlinearity is present within level ranges, interpretation should be based on single level estimates in respect to significance levels. A further potential limitation of this study was the use of a main-effects orthogonal array with zero priors, although a rather advanced D-optimal design was used. Since designing this study, many advances in experimental designs have been made [42]. Modern experimental design approaches may have used this type of design in a pilot, and then utilized priors in, e.g., a Bayesian designs, to improve statistical and respondent efficiency. To prevent possible hypothetical bias and decision heuristics, a scope test for the attributes that used continuous variables like risk would be recommended. Furthermore, a split sample analysis could be used to control for possible biases. As this first pilot test presented here was not intended to test validity and reliability of discrete choice experiment, these were not included, but should be in further research.
Finally, it should be mentioned that to be able to analyze possible interactions, a different experimental design would be needed that allows to test for two-way or all interaction effects. Moreover, to be able to estimate subgroup effects or provide a clustering, further models, such as latent class models, would have to be applied [51]. As it was not the initial aim of the present study to analyze subgroup differences, this will be reported in a future paper. As the main objective of the study was to test the method of DCE for the use in IQWIG context and the data collection was performed in 2010, the presented results are based on the ‘‘old’’ interferon-containing regimes. The results presented herein therefore refer to the treatment properties of the interferon-based therapeutic regime, which should be taken into account before generalization.
Conclusions This pilot study indicates that a DCE can be used to prioritize patient-relevant outcomes from RCTs. The weighting of multiple outcomes allows an indicationspecific and evidence-based measure to be used in health economic evaluations. The transparent weighting of multiple outcomes may appeal to decision-makers, because it can be regarded as a first step to include patient preferences in evidence-based decision-making on the approval and pricing of innovations. The approach applied generally allows for the consideration of patient-relevant trade-offs regarding the benefits and harms of medical interventions. The weighting of patient-relevant outcomes can be derived from multi-criteria decision-making methods such as DCEs, and can be seen as an important step forward. It should be mentioned that patients’ choices among finite sets of treatment options can be interpreted as actions that respond to underlying preferences for outcomes associated to the treatment chosen. Systematic changes in the levels of outcomes within and across alternatives that patients are asked to consider provide the experimental stimuli that reveal the relative importance of each of the patient-relevant outcomes. DCEs rely on the analysis of stated choices for treatment options under experimental conditions to determine the value of the stimuli that drive choices. This advantage also makes preferences elicited through DCEs consistent with the utility-theoretic frameworks used in almost all other areas of economics. Even if no final conclusions can be drawn on the representativeness of our results, the present pilot study indicates that weighted outcomes can be used to support decision-making in health care. With this pilot study, IQWiG is among the first HTA bodies that commissioned exploring methods of preference elicitation in a national
123
A. C. Mu¨hlbacher et al.
context. At present, conjoint analysis and analytic hierarchy process are mentioned in IQWiG’s Methods [52]. As there are diverging views on the status of some methodological questions, for instance, replicability of results, IQWiG encourages continuing efforts in applying this methods. Authors contributions ACM designed the whole study. ACM and SB administered the experiment, analyzed the data and drafted the manuscript. CMD and SB were organizing and coordinating the focus groups. MN analyzed the data and described the analytic model, carried out the stochastic implementation of the regression models. AS and CMD gave technical support during the survey development and commented on the manuscript. JFPB designed the experimental design of the discrete choice experiment and commented on the manuscript. AGG commented on the manuscript in the last phase. Acknowledgments Christin Juhnke, Anika Kaczynski, Andrew Sadler, Fu¨lo¨p Scheibler, and Beate Wiegard provided technical assistance.
8.
9. 10.
11.
12.
13.
14.
15.
16. Compliance with ethical standards Funding This study was funded by the German Institute for Quality and Efficiency in Health Care (IQWiG).
17.
Conflict of interest AS and AGG are employees of IQWiG. CMD was employed by IQWiG until 31 July 2011 and is now employed by Bayer Health Care as well as the Heinrich-Heine University of Du¨sseldorf. AS, AGG, and CMD have received remuneration from IQWiG for their work as external experts. JB is an employee of Johns Hopkins University. MN and AM are employees of Empirical Consulting mbH. AM and SB are employees of the University of Applied Sciences Neubrandenburg.
18.
References 1. Baltussen, R., et al.: Multi-criteria decision analysis to prioritize health interventions: capitalizing on first experiences. Health Policy 96(3), 262–264 (2010) 2. Thokala, P., Duenas A.: Multiple criteria decision analysis for health technology assessment. Value in health, 15 (2012) 3. Rotter, J.S., Foerster, D., Bridges, J.F.: The changing role of economic evaluation in valuing medical technologies. Expert Rev Pharmacoecon Outcomes Res 12(6), 711–723 (2012) 4. Martin, D.K., Giacomini, M., Singer, P.A.: Fairness, accountability for reasonableness, and the views of priority setting decision-makers. Health Policy 61(3), 279–290 (2002) 5. Institute for Quality and Efficiency in Health Care (IQWiG): General Methods, in Version 4.1. 2013, translation based on the German document ‘‘Allgemeine Methoden’’ (Version 4.1): Cologne 6. Danner, M., et al.: Integrating patients’ views into health technology assessment: analytic hierarchy process (AHP) as a method to elicit patient preferences. Int J Technol Assess Health Care 27(04), 369–375 (2011) 7. Institut fu¨r Qualita¨t und Wirtschaftlicheit im Gesundheitswesen (IQWiG): Wahlbasierte Conjoint-Analyse—Pilotprojekt zur Identifikation, Gewichtung und Priorisierung multipler Attribute
123
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
in der Indikation Hepatitis C; IQWiG-Berichte—Nr. 227, unter Mitwirkung von, et al., Editors. 2014: Ko¨ln Thurstone, L.L., A law of comparative judgment. Psychol Rev. 34, 273–286 (1927). (Reprint 1994; Psycholigical Revie, Vol. 101; No. 2; p. 266-270) Lancaster, K.J.: A new approach to consumer theory. J Polit Econ 74(2), 132–157 (1966) McFadden, D., Conditional logit analysis of qualitative choice behavior. In: Zarembka, P. (ed.) Frontiers in Econometrics, pp.105–142. Academic Press, New York (1974) Green, C., Gerard, K.: Exploring the social value of health-care interventions: a stated preference discrete choice experiment. Health Econ 18(8), 951–976 (2009) Baltussen, R., et al.: Priority setting using multiple criteria: should a lung health programme be implemented in Nepal? Health Policy Plan 22(3), 178–185 (2007) de Bekker Grob, E.W., Ryan, M., Gerard, K.: Discrete choice experiments in health economics: a review of the literature. Health Econ 21(2), 145–172 (2010) Bridges, J.F.P., et al.: Conjoint analysis applications in health—a checklist: a report of the ISPOR good research practices for conjoint analysis task force. Value in Health 14(4), 403–413 (2011) Hauber, A.B., et al.: Treatment preferences and medication adherence of people with Type 2 diabetes using oral glucoselowering agents. Diabet. Med 26(4), 416–424 (2009) Mu¨hlbacher, A.C., et al.: Der Wert von Innovationen im Gesundheitswesen: spielen die Patientenpra¨ferenzen eine Rolle? GRPG—R Polit Gesundheit 14(2), 53–62 (2008) Mu¨hlbacher, A.C., Lincke, H.J., Nu¨bling M.: Evaluating patients’ preferences for multiple myeloma therapy, a discrete-choice-experiment. Psycho-soc Med. 5, Doc10 (2008b) Mu¨hlbacher, A.C., et al.: Preferences for treatment of AttentionDeficit/Hyperactivity Disorder (ADHD): a discrete choice experiment. BMC Health Serv Res 9, 149 (2009) Bridges, J.F., et al.: Things are looking up since we started listening to patients: trends in the application of conjoint analysis in health 1982–2007. Patient: Patient-Cent Outcomes Res 14(4), 273–282 (2008) Lancsar, E., Louviere, J.: Conducting discrete choice experiments to inform healthcare decision making: a user’s guide. PharmacoEconomics 26(8), 661–678 (2008) Mu¨hlbacher, A., Bethge, S., Tockhorn, A.: Pra¨ferenzmessung im gesundheitswesen: grundlagen von discrete-choice-experimenten. Gesundheitso¨kon Qualita¨tsmanag 18(4), 159–172 (2013) World Health Organization: Global surveillance and control of hepatitis C. Report of a WHO Consultation organized in collaboration with the Viral Hepatitis Prevention Board. Antwerp, Belgium. J. Viral. Hepat. 6(1), 35–47 (1999) Bayliss, M.S., et al.: A questionnaire to assess the generic and disease-specific health outcomes of patients with chronic hepatitis C. Qual Life Res 7(1), 39–55 (1998) Gutteling, J.J., et al.: Overview of research on health-related quality of life in patients with chronic liver disease. Neth J Med 65(7), 227–234 (2007) Gutteling, J.J., et al.: Determinants of quality of life in chronic liver patients. Aliment Pharmacol Ther 23(11), 1629–1635 (2006) Unal, G., et al.: A psychometric comparison of health-related quality of life measures in chronic liver disease. J Clin Epidemiol 54(6), 587–596 (2001) Younossi, Z.M., et al.: Development of a disease specific questionnaire to measure health related quality of life in patients with chronic liver disease. Gut 45(2), 295–300 (1999) Gralnek, I.M., et al.: Development and evaluation of the Liver Disease Quality of Life instrument in persons with advanced,
Preferences for antiviral therapy of chronic hepatitis C
29. 30.
31. 32.
33. 34.
35.
36.
37.
38.
39.
40.
chronic liver disease-the LDQOL 1.0. Am J Gastroenterol 95(12), 3552–3565 (2000) Moser, G.: Hepatitis C und Psyche. J Gastroenterol Hepatol Erkrank 2(3), 20–22 (2004) Sarrazin, C., et al.: Prophylaxis, diagnosis and therapy of hepatitis C virus (HCV) infection: the German guidelines on the management of HCV infection. Z Gastroenterol 48(2), 289–351 (2010) Treadwell, J.R., Kearney, D., Davila, M.: Health profile preferences of hepatitis C patients. Dig Dis Sci 45(2), 345–350 (2000) Wong, J.B., et al.: Economic and clinical effects of evaluating rapid viral response to peginterferon alfa-2b plus ribavirin for the initial treatment of chronic hepatitis C. Am J Gastroenterol 98(11), 2354–2362 (2003) Craxi, A.: Early virologic response with pegylated interferons. Dig Liver Dis 36(Suppl 3), 340–343 (2004) Backus, L.I., et al.: Pretreatment assessment and predictors of hepatitis C virus treatment in US veterans coinfected with HIV and hepatitis C virus. J Viral Hepat 13(12), 799–810 (2006) Witkos, M., et al.: Predictors of antiviral therapy in a posttransfusion cohort of hepatitis C patients. Can J Gastroenterol 20(2), 107–111 (2006) Cotler, S.J., et al.: Patients’ values for health states associated with hepatitis C and physicians’ estimates of those values. Am J Gastroenterol 96(9), 2730–2736 (2001) Fraenkel, L., et al.: Patients’ preferences for treatment of hepatitis C. Med Decis Mak: Int J Soc Med Decis Mak 30(1), 45–57 (2010) Schackman, B.R., et al.: Quality-of-life tradeoffs for hepatitis C treatment: do patients and providers agree? Med Decis Mak 28(2), 233–242 (2008) Sawtooth Software: The CBC system for choice-based conjoint analysis. In: CBC—Technical Paper (ed.) Technical Paper Series, Sawtooth Software, Inc., Washington (2008) Ryan, M., Gerard, K.: Using discrete choice experiments to value health care programmes: current practice and future research reflections. Appl Health Econ Health Policy 2(1), 55–64 (2003)
41. Street, D., Burgess, L.: The construction of optimal stated choice experiments: theory and methods. Wiley-Interscience Series in Probability and Statistics, New Jersey (2007) 42. Johnson, R.F., et al.: Constructing experimental designs for discrete-choice experiments: report of the ISPOR conjoint analysis experimental design good research practices task force. Value In Health 16(1), 3–13 (2013) 43. Orme, B.K., Sample size issues for conjoint analysis. In: Getting Started with Conjoint Analysis, 2nd edn. pp. 57–66, Research Publishers LLC, Madison (2010) 44. Bech, M., Kjaer, T., Lauridsen, J.: Does the number of choice sets matter? Results from a web survey applying a discrete choice experiment. Health Econ 20(3), 273–286 (2011) 45. Brett Hauber, A., et al.: Patient preferences and assessment of likely adherence to hepatitis C virus treatment. J Viral Hepatitis 18(9), 619–627 (2011) 46. Kauf, T.L., et al.: Patients’ willingness to accept the risks and benefits of new treatments for chronic hepatitis C virus infection. Patient: Patient-Cent Outcomes Res 5(4), 265–278 (2012) 47. Manns, M.P., Cornberg, M.: Sofosbuvir: the final nail in the coffin for hepatitis C? Lancet Infect Dis 13(5), 378–379 (2013) 48. Schneider, M., Zeuzem, S., Sarrazin, C.: Aktueller Stand der Hepatitis-C-Therapie. Der Gastroenterologe 8(1), 27–34 (2013) 49. Koch-Institut, Robert: Zur Situation bei wichtigen Infektionskrankheiten in Deutschland: virushepatitis B, C und D im Jahr 2010. Epidemiol Bull 29, 261–274 (2010) 50. Shepard, C.W., et al.: Cost-effectiveness of conjugate meningococcal vaccination strategies in the United States. Pediatrics 115(5), 1220–1232 (2005) 51. Ryan, M., Gerard, K., Amaya-Amaya M., Using discrete choice experiments to value health and health care. In: The Economics of Non-Market Goods and Resources, 11, p. XIX 254 S, Springer, Dordrecht 2008 52. Institut fu¨r Qualita¨t und Wirtschaftlicheit im Gesundheitswesen (Institute for Quality and Efficiency in Health Care) (IQWiG), Allgemeine Methoden 4.2. Ko¨ln (2015)
123
ORIGINAL PAPER
Preferences for antiviral therapy of chronic hepatitis C: a discrete choice experiment Axel C. Mu¨hlbacher1,2 • John F. P. Bridges3 • Susanne Bethge1 • Ch.-Markos Dintsios5,6 • Anja Schwalm4 • Andreas Gerber-Grote4 • Matthias Nu¨bling7
Received: 19 March 2015 / Accepted: 11 January 2016 Ó Springer-Verlag Berlin Heidelberg 2016
Abstract Background The German Institute for Quality and Efficiency in Health Care (IQWiG) uses patient-relevant outcomes to inform decision-makers. Objective IQWiG conducted a pilot study to examine whether discrete choice experiments (DCEs) can be applied in health economic evaluations in Germany to identify, weight, and prioritize multiple patient-relevant outcomes, using the example of antiviral therapy for chronic hepatitis C (HCV). A further objective was to contribute to a more structured approach towards eliciting and comparing preferences across key stakeholders. Methods In autumn 2010, a DCE questionnaire was sent to patients with chronic HCV to estimate preferences across seven outcomes (‘‘attributes’’), including treatment efficacy [sustained viral response (SVR) at 6 months],
& Axel C. Mu¨hlbacher [email protected]
adverse effects (flu-like symptoms, gastrointestinal symptoms, psychiatric symptoms, and skin symptoms/alopecia), and measures of treatment burden (duration of therapy, frequency of injections). A linear model and an effects coded full model were applied to assess the relative importance of the attributes. Results In total N = 326 patients were included. A clear preference for SVR was shown; frequency of injections and duration of therapy shared the second rank, while psychiatric symptoms ranked third. The duration of flu-like symptoms was the least important attribute. Conclusion Our findings indicate that it is possible to perform a DCE at the national level in a health technology assessment agency. The weighting of multiple outcomes allows an indication-specific and evidence-based measure to be used in health economic evaluations. In decisionmaking in health care, the approach generally allows for consideration of patient-relevant trade-offs regarding the benefits and harms of medical interventions. Keywords Discrete choice experiment (DCE) Conjoint analysis (CA) Hepatitis C virus (HCV) Health technology assessment (HTA) Patient preferences Priority setting
1
Institute for Health Economics and Health Care Management (IGM), University of Applied Sciences Neubrandenburg, Brodaer Straße 2, 17033 Neubrandenburg, Germany
2
Center for Health Policy and Inequalities Research, Duke Global Health Institute, Duke University, Durham, NC, USA
3
Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA
JEL Classification
4
Institute for Quality and Efficiency in Health Care (IQWiG), Cologne, Germany
Introduction
5
Strategic Market Access Intelligence, Bayer Health Care, Leverkusen, Germany
6
Department of Public Health, Faculty of Medicine, Heinrich-Heine University, Du¨sseldorf, Germany
7
Empirical Consulting mbH (GEB mbH), Freiburg, Germany
Health policy decision-makers and clinicians largely draw their conclusions on the benefits and harms of medical interventions on the basis of different outcomes investigated in clinical trials. However, due to the large number and diversity of these outcomes, it is a highly complex task
I 11 H 51 I 18
123
A. C. Mu¨hlbacher et al.
to determine the true ‘‘value’’ of competing interventions. Even if specific decision criteria are established, their relative importance to patients and other stakeholders remains unclear. The use of multiple decision criteria therefore calls for the weighing of benefits and harms to inform decisionmaking [1]. A health technology assessment (HTA) using multiple outcomes from a clinical trial can thus only provide comprehensive information for decision-making if these outcomes are weighted and can be reduced to a common denominator. Various methods of multi-criteria decision-making have been tested in the past years, such as the analytical hierarchy process (AHP) and a stated preference method called, ‘‘conjoint analysis’’, including its choice-based form, the discrete choice experiment (DCE) [2]. However, no established scientific standard for the identification, weighting and prioritization of outcomes currently exists. It is therefore unclear how multiple outcomes should be handled in the decision-making process. To prepare guidance for transparent prioritization and allocation decisions, it is desirable to determine the actual value judgments of key stakeholders. In recent years, many HTA agencies have aimed to broaden their scope of evaluation by including a wide range of perspectives, including the patient perspective [3]. This approach is considered to be an important element in transparent priority setting [4]. However, up to now there has been no practical guidance on how to involve key stakeholders in HTA. The Institute for Quality and Efficiency in Health Care (IQWiG), the main German HTA agency, assesses the benefits of interventions for patients on the basis of patientrelevant outcomes. According to the German Social Code Book V (SGB V), the outcomes primarily considered are mortality, morbidity and health-related quality of life [5]. IQWiG must ensure that the assessment of the patientrelevant benefit of a medical intervention is conducted according to internationally accepted standards of evidence-based medicine and health economics [§139a (4) SGB V]. In Germany, the health economic evaluation of an intervention is supposed to be based directly on the patientrelevant outcomes of clinical trials; therefore, various methods for weighting outcomes had to be tested. In addition, IQWiG and other organizations have been promoting the position that elicitation of patient preferences is one of the most important challenges in the construction of a utility metric used in a cost-benefit framework. IQWiG therefore initiated a pilot study to examine whether a DCE can be applied in health economic evaluations in Germany to identify, weight and prioritize multiple patient-relevant outcomes, using the example of antiviral therapy for chronic hepatitis C (HCV).1 This study focused on the valuation of different decision criteria or patient relevant endpoints, but did not assess quality of life
123
or health episodes. The results of the study were also to be used to inform regulatory and health policy decisionmakers about the feasibility of the approach in the context of HTA and about the design of surveys on patient preferences. A further objective was to contribute to a more structured approach towards eliciting and comparing preferences across key stakeholders. The present pilot study was conducted in the form of a survey and in parallel to an AHP study in patients with depression [6]. The full German-language report and an English-language summary are available on the IQWiG website [7].
Methods Choice of approach In contrast to AHP methods [6], DCEs can be used to compute relative importance weights for various treatment attributes that influence the benefit for patients. DCEs are grounded in psychology [8], economics [9], and econometrics [10]. They are based on respondents’ assessments of competing alternatives within choice sets. Over the last decade, DCEs have been used to analyze respondents’ perspectives on resource allocation decisions [11–13], as well as to determine the preferences of patients, insured persons or experts in the healthcare system [14–19]. Medical interventions can be described by means of characteristics or outcomes of a treatment, which are referred to as attributes, and by the levels of these attributes. In a DCE, utility is measured on an arbitrary scale, and the overall importance of each attribute depends on the range of attribute levels chosen for the experiment. Several guidelines have been developed for the application of DCEs in health care [14, 20, 21]. Context of disease The World Health Organization (WHO) estimates that the global incidence of HCV is 3 % [22]. The disease leads to increased mortality and morbidity rates, as well as to a significant impairment in quality of life [23–28]. In 80 % of the cases, the course of disease is chronic [29]. When the pilot study was performed in 2010, the clinical guidelines 1 Note The present pilot study on the treatment of HCV uses attributes and outcomes that might not necessarily correspond to patient-relevant outcomes pursuant to SGB V (i.e., outcomes describing morbidity, mortality, and health-related quality of life). It cannot be concluded from the language used in the present paper that IQWiG would regard the attributes and outcomes applied to represent (patient-relevant) outcomes in the event of a benefit assessment.
Preferences for antiviral therapy of chronic hepatitis C
for the treatment of chronic HCV recommended a combination therapy with peg-interferon and ribavirin for 48 or 72 weeks; telaprevir and boceprevir, as well as interferonfree regimens, were not available at that time [30]. The clinical objective of antiviral HCV therapy is to achieve sustained virological response (SVR), i.e., permanent eradication of the virus. Treatment decisions in patients with HCV should generally be made taking patient preferences into account [31–38]. Besides SVR, further patientrelevant evaluation criteria should therefore be considered. Patient recruitment and selection The survey was conducted in September and October 2010 with paper-and-pencil questionnaires. Questionnaires were sent to patients with HCV who were members of the ‘‘Deutsche Leberhilfe e.V.’’ (German Liver Support Group), a large patient organization. Alternatively, they could fill out an online questionnaire provided by the same organization. Patients with HCV who were older than 18 years of age were eligible for inclusion. Patients with active alcohol or drug abuse or co-infection with human immunodeficiency virus were excluded because of possible interferences with medication. To achieve a sufficient sample size, no HCV genotype was excluded. Ethical approval All respondents were informed in writing about the study objective, as well as the potential benefits and risks, prior to participation and provided informed consent. Participation was voluntary and not reimbursed. The study, together with all information materials and the questionnaire, was ethically approved by the Medical Association of Northern Rhine. Survey components The questionnaire was structured as follows: Section 1: patient characteristics The first section contained questions on general sociodemographics (e.g., gender, age, marital status, and educational level), as well as medical history (e.g., current health status, year of first diagnosis, current treatment status, and genotype). Section 2: attributes and levels The second section contained an introduction to the attributes and levels used in the survey. To ensure that this selection was based on the best possible evidence, a
targeted literature search was conducted in March and April 2010 in the PUBMED, MEDLINE, SPRINGER, and THIEME databases to identify the most important attributes and levels. The search covered the relevance and effects of chronic HCV, as well as treatment options. The main focus of the literature search was on randomized controlled trials (RCTs) and evidence-based guidelines. In addition, two focus groups (n = 9) were formed to ensure that no relevant outcomes had been overlooked, and to examine the possible level ranges, as well as the comprehensibility of the attributes. At a later stage, the focus groups were used for the qualitative validation of the selected attributes and their values (levels) to ensure a comprehensive account of the specific assumptions of the affected patients. On the basis of the results of the literature search and the focus groups, seven attributes were extracted and included in the questionnaire. The above nine subjects pre-tested a draft version of the questionnaire, which was found to be manageable. The final attributes and levels included in the questionnaire and the underlying literature are shown in Table 1. Section 3: choice tasks The third section contained the choice tasks (see Fig. 1). Patient preferences were measured by asking which of two hypothetical treatment options respondents would choose. Within the choice tasks, they were not required to indicate the strength of their preferences or to assess the confidence they had in their choice. In order to clarify the meaning of the attributes and to ensure that these attributes were understood in the same way, an easy-to-understand, non-technical explanation was provided for each attribute. The full profile approach was applied for the construction of the choice tasks, which means that each choice set included all seven attributes. In DCE research, four to eight attributes per choice set are seen as appropriate [39, 40]. As clinical guidelines strongly recommend treatment of HCV, no ‘‘opt-out’’ or ‘‘status-quo’’ alternative was included in the choice tasks. After the choice tasks, the respondents could make their own comments and were asked to indicate how difficult it was to fill in the questionnaire. Experimental design Given that the study did not aim to explore any specific interactions, a main-effects experimental design was used. Specifically, we used an orthogonal experimental design that accounted for three attribute levels by applying a 6^7 orthogonal array, producing a D-optimal experiment [41].
123
A. C. Mu¨hlbacher et al. Table 1 Attributes and levels included in the questionnaire
Attribute Duration of antiviral therapy
Level
References
12 weeks
[29–30]
24 weeks 48 weeks Frequency of interferon injections
Three times per week
[29–30]
Once per week Once every 14 days Duration of flu-like symptoms after injection
1 day after injection 2 days after injection 3 days after injection
Probability of gastro-intestinal symptoms
25 of 100 persons (25 %) 35 of 100 persons (35 %) 45 of 100 persons (45 %)
Probability of psychiatric symptoms
35 of 100 persons (35 %) 45 of 100 persons (45 %)
Probability of skin symptoms and/or alopecia
55 of 100 persons (55 %) 35 of 100 persons (35 %) 45 of 100 persons (45 %) 55 of 100 persons (55 %)
Probability of SVR 6 months after end of therapy
45 of 100 persons (45 %) 55 of 100 persons (55 %) 65 of 100 persons (65 %)
Fig. 1 Example of a choice task
Therapy A
Therapy B
24 weeks
48 weeks
Frequency of interferon injections
once every 2 weeks
once a week
Duration of flu-like symptoms after injection
3 days after injection
one day after injection
Probability of gastrointestinal symptoms
25 out of 100 people (25%)
45 out of 100 people (45%)
Probability of psychiatric symptoms
55 out of 100 people (55%)
45 out of 100 people (45%)
Probability of skin problems or alopecia
55 out of 100 people (55%)
45 out of 100 people (45%)
Probability of SVR 6 months after treatment
55 out of 100 people (55%)
45 out of 100 people (45%)
□
□
Duration of treatment
Please choose one therapy option.
The approach (i.e., use of a higher-order design to generate both profiles in the pair simultaneously) was considered a more efficient, and potentially more accurate, approach to construct the orthogonal experimental design in the absence of priors [42]. The planned sample size according to the Orme calculation was a minimum of n = 168 participants (n = 42 per DCE Block) [43].
123
The experimental design led to 72 paired-comparison choice tasks, which were randomly assigned to four blocks of 18 choice tasks each, in order to reduce the cognitive burden for patients. The final experimental design resulted in four different versions of the questionnaire; each respondent was randomly allocated to one of these versions. All choice tasks were forced choice options, i.e.,
Preferences for antiviral therapy of chronic hepatitis C
respondents had to choose one option. They were asked to make 18 choices, as other studies have shown that this number of choices is manageable without significant differences in results [44]. The whole experimental design and all choice-set are available at: https://www.iqwig.de/down load/GA10-03_Arbeitspapier_Version_1-1_Conjoint-Ana lyse-Pilotprojekt.pdf. The testing of the experimental design confirmed 100 % D-optimality for main effects in a paired-comparison experiment, as well as orthogonality at both the profile and experiment level. Furthermore, level balance (i.e., each level was equally likely to appear on the first or second profile in the experiment) and zero overlap (occurrence of the same level in one choice task) were given. Data analysis In order to assess the relative importance of attributes, we fitted a linear model, as well as an effects coded model (full model), using SPSS 18 and STATA 11. Within the linear model, the difference between levels was coded in only one variable for each attribute. All three levels are thereby treated as linear degrees with identical distances between levels. In the full model, effects coding was used for all seven attributes. The worst level (e.g., highest probability of adverse effects) was in each case (model) coded as the reference category. Subsequently, the difference between treatment options A and B was generated and coded in two variables for the upper and the middle level. The linearity assumption of the levels underlining the linear model needs to be tested in the full model showing the estimated preference weights for all levels included. If one assumes a priori that the levels are not nominally, but ordinally scaled, it can be tested in the full model whether the linearity assumption applies across all levels of attributes. The coefficient per attribute therefore indicates the increase per level (difference between A and B). Linearity must be tested, because identical distances between levels cannot be presumed. If linearity is confirmed, the linear model suffices for the identification and weighting of attributes, as this model is adequate to describe linear relationships. In our study, we calculated to what extent level 1 takes the middle ranking in preference between levels -1 and 1. For the effects coding used, this means that the estimated coefficient of the middle level should be close to zero. Quality of the statistical model The quality of the various models was assessed and compared using the Akaike Information Criterion and Bayesian Information Criterion, which are both used to assess the general measurement quality of models by means of the likelihood ratio test. Model quality can also be assessed by
means of the proportion of choices correctly classified by the model (proportion of choice tasks with a predicted probability of choice of [50 % in the model compared with the actual choices made by patients). The full model and the linear model correctly predicted 74.24 and 74.11 % of the choices, respectively. The linear model could therefore be applied. Data interpretation Within the estimated models, the coefficients presented indicate the relative importance of an attribute, e.g., the level within the treatment choice. The higher the value of a coefficient, the higher the importance of the related attribute/level within the decision. A positive coefficient indicates that the attribute is valuable to the patient. A negative coefficient indicates that the participant attaches a negative value. The final ranking is based on the coefficient for each patient-relevant outcome out of the coefficients in the linear model. The coefficient can be interpreted as the relative value of an improvement of one level for each attribute. The full model was effect coded, and therefore displays coefficients for each single level per attribute (Table 3). We generated the mean relative importance of each attribute by computing the differences between the highest and lowest coefficients for the levels of each attribute. We then normalized the scale by assigning ten units to the most important attribute and measured the other attributes’ importance relative to this change. The mean relative importance score for each attribute expresses an improvement for an attribute on a scale from 0 (worst level) to 10 (best level); Fig. 2.
Results Respondent characteristics A total of 326 patients, 64 % of them female, participated in the survey (234 completed the paper and 92 the online version. Controlling for the mode of administration (paper or online version), no significant differences were detected.) Their mean age was 58.9 years. Within the study sample 39.9 % had already completed treatment, 7.7 % were currently undergoing treatment, 24 % had discontinued, and 28.2 % had not yet started a HCV treatment. 69.9 % were genotype 1. 60.5 % stated that it was easy or very easy to fill in the survey. The demographic and clinical data are shown in Table 2. Of the 326 questionnaires returned, 17 contained no answers to the choice tasks and could not be analyzed. In all, 5252 decisions from 309 respondents were used for the final analysis.
123
A. C. Mu¨hlbacher et al. Fig. 2 Graphic display of level estimates and normalization of coefficients
Results of the preference elicitation models The analysis of the linear model resulted in the following weights for the seven attributes: ‘‘SVR 6 months after end of therapy’’ (coef.: 0.8041), ‘‘frequency of interferon injections’’ (coef.: 0.2966), ‘‘duration of antiviral therapy’’ (coef.: 0.2502), ‘‘probability of psychiatric symptoms’’ (coef.: 0.1857), ‘‘probability of gastrointestinal symptoms’’ (coef.: 0.1233), ‘‘probability of skin problems or alopecia’’ (coef.: 0.1054), and ‘‘duration of flu-like symptoms after injection’’ (coef.: 0.1052); Table 3.
123
On the basis of the normalized difference between the coefficients of the best and worst level of the full model, the attributes ranked as follows: SVR (10), frequency of injections (3.69), duration of therapy (3.18), psychiatric symptoms (2.29), skin symptoms/ alopecia (1.48), gastrointestinal symptoms (1.44), and flu-like symptoms (1.27); see Fig. 2. If confidence intervals were taken into account, SVR ranked first, with frequency of injections and duration of therapy sharing the second rank, while psychiatric symptoms ranked third.
Preferences for antiviral therapy of chronic hepatitis C
Results of the significance tests
Table 2 Patient characteristics Item
N
N in %
Overall
326
100
Gender Female
208
63.8
Male
117
35.9
Not reported
1
0.31
Age Median (years) Not reported
58
–
7
2.15
245
75.15
20
6.13
Marital status Married/living with a partner Single Separated
5
1.53
Widowed
16
4.91
Divorced
38
11.66
2
0.61
Not reported
In the comparison of attributes, the preference for SVR was significantly higher than for all other attributes. The difference between the coefficients of ‘‘duration of therapy’’ and ‘‘frequency of injections’’ was not significant, but was significant between these two attributes and all other ones. These two attributes were therefore ranked in second place. The attribute ‘‘psychiatric symptoms’’ ranked in fourth place. ‘‘Gastrointestinal symptoms’’ (fifth place) and ‘‘skin symptoms/alopecia’’ were weighted higher than ‘‘flu-like symptoms.’’ As no significant differences were shown between these three attributes (overlapping confidential intervals), they therefore shared a conjoint fifth to seventh place.
Discussion
Net-income/pension in last month Less than 1000 Euro
108
33.13
1000–2000 Euro
124
38.04
2000–3000 Euro
49
15.03
3000–4000 Euro
12
3.68
More than 4000 Euro
15
4.60
Not reported
18
5.52
No fibrosis
67
20.55
Minor fibrosis
80
24.54
Moderate fibrosis
69
21.17
Current status of disease
Comparison with previous research
Severe fibrosis
12
3.68
Cirrhosis
50
15.34
Not known
40
12.27
8
2.45
Type I
228
69.94
Type II
30
9.20
Type III
27
8.28
Type IV
9
2.76
Type V
3
0.92
22
6.75
7
2.15
Not reported HCV genotype
Not known Not reported Diagnosis of hepatocellular carcinoma Yes
3
0.92
293
89.88
29
8.90
1
0.31
No treatment started
92
28.22
Currently being treated
25
7.67
130
39.88
78
23.93
1
0.31
No Not known Not reported Current therapy status
Treatment completed Treatment discontinued Not reported
In a sample of more than 300 patients, patient preferences were determined in a DCE for attributes of antiviral HCV therapy. Our study shows that patients are able to consider alternative scenarios for treatments and outcomes when choosing their optimum treatment strategy.
The results of the present study are in line with similar analyses in other healthcare systems. For example, a preference study by Fraenkel et al. [37] conducted in the US also found that achieving SVR 6 months after the end of therapy was the most important attribute of anti-HCV therapy. Moreover, the study showed that the decision to treat depends on the severity of potential treatment-related adverse effects. If adverse effects were expected to be mild, 67 % of patients reported that they would opt for treatment. However, if adverse effects were expected to be severe, this rate decreased to 51 %. Two further US studies (Hauber et al. [45] and Kauf et al. [46]) also found SVR to be the most important attribute for patients within a treatment decision. As the shared second rank shows, the attributes ‘‘frequency of injections’’ and ‘‘duration of therapy’’ are essential outcomes in the assessment of the benefits of antiviral HCV therapy by patients. Treatment regimens usually last between 48 and 72 weeks, and daily interferon injections seem to represent a great burden and should be taken into account in a benefit assessment [30, 47, 48]. The same applies to all side effect attributes, which should also be taken into consideration. As all attributes resulted in significant coefficients, all are relevant for patients and need to be addressed within value judgments.
123
A. C. Mu¨hlbacher et al. Table 3 Parameters estimated from the linear and full model Attributes and levels
Duration of antiviral therapy
Linear model
Full model
Coefficients
Odds ratio
SE coeff
95 % confidence intervals
0.2502
1.2843
0.0234
0.204
p
Coefficients
Odds ratio
SE coeff
95 % confidence intervals
p
–0.356 –0.248 \0.001
0.296 \0.001
48 weeks
–0.3019
0.7394
0.0270
24 weeks
0.0869
1.0907
0.0274
0.030
0.140 \0.01
12 weeks
0.2151
1.2399
0.0275
0.160
0.270 \0.001
–0.3748 0.1490
0.6874 1.1607
0.0275 0.0276
0.2258
1.2533
0.0273
Frequency of interferon injections
0.2966
1.3453
0.0234
0.251
0.342 \0.001
Three times per week Once per week Once per 14 days Duration of flu-like symptoms after injection
0.1052
1.1109
0.0232
0.060
–0.429 –0.321 \0.001 0.090 0.200 \0.001 0.170
0.280 \0.001
0.151 \0.001
3 days after injection
–0.0877
0.9160
0.0269
–0.140 –0.035 \0.001
2 days after injection
–0.0319
0.9686
0.0277
–0.090
1 day after injection
0.1196
1.1270
0.0273
0.070
45 of 100 persons
–0.1296
0.8785
0.0270
–0.183 –0.077 \0.001
35 of 100 persons
0.0250
1.0253
0.0275
–0.030
25 of 100 persons
0.1046
1.1103
0.0274
0.050
Probability of gastrointestinal symptoms
Probability of psychiatric symptoms
0.1233
0.1857
1.1312
1.2041
0.0233
0.0234
0.078
0.140
0.020
n.s.
0.170 \0.001
0.169 \0.001
0.080
n.s.
0.160 \0.001
0.232 \0.001
55 of 100 persons
–0.1820
0.8336
0.0277
–0.236 –0.128 \0.001
45 of 100 persons
–0.0086
0.9914
0.0272
–0.060
35 of 100 persons
0.1906
1.2100
0.0273
0.140
55 of 100 persons
–0.1509
0.8599
0.0274
45 of 100 persons 35 of 100 persons
0.0907 0.0602
1.0949 1.0620
0.0271 0.0268
45 of 100 persons
–0.8233
0.4390
0.0292
–0.881 –0.766 \0.001
55 of 100 persons
0.0192
1.0194
0.0256
–0.031
65 of 100 persons
0.8041
2.2347
0.0296
0.750
Probability of skin symptoms and/or alopecia
Probability of SVR 6 months after therapy
0.1054
0.8041
1.1112
2.2347
0.0263
0.0261
0.060
0.753
0.040
n.s.
0.240 \0.001
0.151 \0.001
–0.205 –0.097 \0.01 0.040 0.010
0.140 \0.01 0.110 \0.05
0.855 \0.001
0.069
n.s.
0.860 \0.001
# obs.: 5252; N = 326; LR chi2(7) = 1563.26; Log likelihood = –2852.7476; Prob [ chi2 = 0.0000; # obs.: 5252; N = 326
Importance of the study and research context When confronted with a multi-criteria decision problem (as is the case in the treatment of HCV), it needs to be taken into account that the value of (clinical) outcomes for the patient should be considered, not the actual clinical result, as has been common practice thus far. Measurability must thus be assumed not only for the difference in clinical
123
results, but also for the difference in values between the outcome levels. In the context of an allocation decision, it is important that the values of the outcomes are used. The benefit of a medical intervention for patients can only be assessed with the involvement of the affected patients themselves. As a one-dimensional indicator to explain choices about decisions, patient preferences represent the extent to which a therapy alternative should be
Preferences for antiviral therapy of chronic hepatitis C
favored from the patient perspective. The perceived or expected benefit (or harm) from this perspective is thereby the basis for patient preferences and consequently one of the explanatory factors for a patient’s choice of action. For this reason, on the basis of the known preferences of an individual or a population of patients, conclusions can be drawn about the relative benefit of a treatment alternative. Within the legal frameworks of various jurisdictions and decision-making bodies, results of such studies could feed into decisions on reimbursement and pricing of interventions. Limitations of the study With regard to the generalizability of the results, it should be noted that the study sample was characterized by strong variations of individual characteristics, even if not fully representative. It was aimed to include a broad range of hepatitis patients to be able to present a most informative model. The mean age of patients was nearly 60 years and nearly two-thirds were female; however, in Germany patients with hepatitis C are generally younger and predominantly male [49, 50]. This lack of representativeness may, among other things, be explained by the fact that patient recruitment was conducted solely via one patient organization, meaning that not all patients eligible (e.g., those in prison or other high-risk patients) could be reached. Other preference studies have shown that values of patients may depend on their cultural background and the healthcare system they live in [46]. This should also be taken into account when interpreting the study results, as the study was conducted in Germany. While interpreting the data, the linearity over the range of levels evaluated needs to be verified. If nonlinearity is present within level ranges, interpretation should be based on single level estimates in respect to significance levels. A further potential limitation of this study was the use of a main-effects orthogonal array with zero priors, although a rather advanced D-optimal design was used. Since designing this study, many advances in experimental designs have been made [42]. Modern experimental design approaches may have used this type of design in a pilot, and then utilized priors in, e.g., a Bayesian designs, to improve statistical and respondent efficiency. To prevent possible hypothetical bias and decision heuristics, a scope test for the attributes that used continuous variables like risk would be recommended. Furthermore, a split sample analysis could be used to control for possible biases. As this first pilot test presented here was not intended to test validity and reliability of discrete choice experiment, these were not included, but should be in further research.
Finally, it should be mentioned that to be able to analyze possible interactions, a different experimental design would be needed that allows to test for two-way or all interaction effects. Moreover, to be able to estimate subgroup effects or provide a clustering, further models, such as latent class models, would have to be applied [51]. As it was not the initial aim of the present study to analyze subgroup differences, this will be reported in a future paper. As the main objective of the study was to test the method of DCE for the use in IQWIG context and the data collection was performed in 2010, the presented results are based on the ‘‘old’’ interferon-containing regimes. The results presented herein therefore refer to the treatment properties of the interferon-based therapeutic regime, which should be taken into account before generalization.
Conclusions This pilot study indicates that a DCE can be used to prioritize patient-relevant outcomes from RCTs. The weighting of multiple outcomes allows an indicationspecific and evidence-based measure to be used in health economic evaluations. The transparent weighting of multiple outcomes may appeal to decision-makers, because it can be regarded as a first step to include patient preferences in evidence-based decision-making on the approval and pricing of innovations. The approach applied generally allows for the consideration of patient-relevant trade-offs regarding the benefits and harms of medical interventions. The weighting of patient-relevant outcomes can be derived from multi-criteria decision-making methods such as DCEs, and can be seen as an important step forward. It should be mentioned that patients’ choices among finite sets of treatment options can be interpreted as actions that respond to underlying preferences for outcomes associated to the treatment chosen. Systematic changes in the levels of outcomes within and across alternatives that patients are asked to consider provide the experimental stimuli that reveal the relative importance of each of the patient-relevant outcomes. DCEs rely on the analysis of stated choices for treatment options under experimental conditions to determine the value of the stimuli that drive choices. This advantage also makes preferences elicited through DCEs consistent with the utility-theoretic frameworks used in almost all other areas of economics. Even if no final conclusions can be drawn on the representativeness of our results, the present pilot study indicates that weighted outcomes can be used to support decision-making in health care. With this pilot study, IQWiG is among the first HTA bodies that commissioned exploring methods of preference elicitation in a national
123
A. C. Mu¨hlbacher et al.
context. At present, conjoint analysis and analytic hierarchy process are mentioned in IQWiG’s Methods [52]. As there are diverging views on the status of some methodological questions, for instance, replicability of results, IQWiG encourages continuing efforts in applying this methods. Authors contributions ACM designed the whole study. ACM and SB administered the experiment, analyzed the data and drafted the manuscript. CMD and SB were organizing and coordinating the focus groups. MN analyzed the data and described the analytic model, carried out the stochastic implementation of the regression models. AS and CMD gave technical support during the survey development and commented on the manuscript. JFPB designed the experimental design of the discrete choice experiment and commented on the manuscript. AGG commented on the manuscript in the last phase. Acknowledgments Christin Juhnke, Anika Kaczynski, Andrew Sadler, Fu¨lo¨p Scheibler, and Beate Wiegard provided technical assistance.
8.
9. 10.
11.
12.
13.
14.
15.
16. Compliance with ethical standards Funding This study was funded by the German Institute for Quality and Efficiency in Health Care (IQWiG).
17.
Conflict of interest AS and AGG are employees of IQWiG. CMD was employed by IQWiG until 31 July 2011 and is now employed by Bayer Health Care as well as the Heinrich-Heine University of Du¨sseldorf. AS, AGG, and CMD have received remuneration from IQWiG for their work as external experts. JB is an employee of Johns Hopkins University. MN and AM are employees of Empirical Consulting mbH. AM and SB are employees of the University of Applied Sciences Neubrandenburg.
18.
References 1. Baltussen, R., et al.: Multi-criteria decision analysis to prioritize health interventions: capitalizing on first experiences. Health Policy 96(3), 262–264 (2010) 2. Thokala, P., Duenas A.: Multiple criteria decision analysis for health technology assessment. Value in health, 15 (2012) 3. Rotter, J.S., Foerster, D., Bridges, J.F.: The changing role of economic evaluation in valuing medical technologies. Expert Rev Pharmacoecon Outcomes Res 12(6), 711–723 (2012) 4. Martin, D.K., Giacomini, M., Singer, P.A.: Fairness, accountability for reasonableness, and the views of priority setting decision-makers. Health Policy 61(3), 279–290 (2002) 5. Institute for Quality and Efficiency in Health Care (IQWiG): General Methods, in Version 4.1. 2013, translation based on the German document ‘‘Allgemeine Methoden’’ (Version 4.1): Cologne 6. Danner, M., et al.: Integrating patients’ views into health technology assessment: analytic hierarchy process (AHP) as a method to elicit patient preferences. Int J Technol Assess Health Care 27(04), 369–375 (2011) 7. Institut fu¨r Qualita¨t und Wirtschaftlicheit im Gesundheitswesen (IQWiG): Wahlbasierte Conjoint-Analyse—Pilotprojekt zur Identifikation, Gewichtung und Priorisierung multipler Attribute
123
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
in der Indikation Hepatitis C; IQWiG-Berichte—Nr. 227, unter Mitwirkung von, et al., Editors. 2014: Ko¨ln Thurstone, L.L., A law of comparative judgment. Psychol Rev. 34, 273–286 (1927). (Reprint 1994; Psycholigical Revie, Vol. 101; No. 2; p. 266-270) Lancaster, K.J.: A new approach to consumer theory. J Polit Econ 74(2), 132–157 (1966) McFadden, D., Conditional logit analysis of qualitative choice behavior. In: Zarembka, P. (ed.) Frontiers in Econometrics, pp.105–142. Academic Press, New York (1974) Green, C., Gerard, K.: Exploring the social value of health-care interventions: a stated preference discrete choice experiment. Health Econ 18(8), 951–976 (2009) Baltussen, R., et al.: Priority setting using multiple criteria: should a lung health programme be implemented in Nepal? Health Policy Plan 22(3), 178–185 (2007) de Bekker Grob, E.W., Ryan, M., Gerard, K.: Discrete choice experiments in health economics: a review of the literature. Health Econ 21(2), 145–172 (2010) Bridges, J.F.P., et al.: Conjoint analysis applications in health—a checklist: a report of the ISPOR good research practices for conjoint analysis task force. Value in Health 14(4), 403–413 (2011) Hauber, A.B., et al.: Treatment preferences and medication adherence of people with Type 2 diabetes using oral glucoselowering agents. Diabet. Med 26(4), 416–424 (2009) Mu¨hlbacher, A.C., et al.: Der Wert von Innovationen im Gesundheitswesen: spielen die Patientenpra¨ferenzen eine Rolle? GRPG—R Polit Gesundheit 14(2), 53–62 (2008) Mu¨hlbacher, A.C., Lincke, H.J., Nu¨bling M.: Evaluating patients’ preferences for multiple myeloma therapy, a discrete-choice-experiment. Psycho-soc Med. 5, Doc10 (2008b) Mu¨hlbacher, A.C., et al.: Preferences for treatment of AttentionDeficit/Hyperactivity Disorder (ADHD): a discrete choice experiment. BMC Health Serv Res 9, 149 (2009) Bridges, J.F., et al.: Things are looking up since we started listening to patients: trends in the application of conjoint analysis in health 1982–2007. Patient: Patient-Cent Outcomes Res 14(4), 273–282 (2008) Lancsar, E., Louviere, J.: Conducting discrete choice experiments to inform healthcare decision making: a user’s guide. PharmacoEconomics 26(8), 661–678 (2008) Mu¨hlbacher, A., Bethge, S., Tockhorn, A.: Pra¨ferenzmessung im gesundheitswesen: grundlagen von discrete-choice-experimenten. Gesundheitso¨kon Qualita¨tsmanag 18(4), 159–172 (2013) World Health Organization: Global surveillance and control of hepatitis C. Report of a WHO Consultation organized in collaboration with the Viral Hepatitis Prevention Board. Antwerp, Belgium. J. Viral. Hepat. 6(1), 35–47 (1999) Bayliss, M.S., et al.: A questionnaire to assess the generic and disease-specific health outcomes of patients with chronic hepatitis C. Qual Life Res 7(1), 39–55 (1998) Gutteling, J.J., et al.: Overview of research on health-related quality of life in patients with chronic liver disease. Neth J Med 65(7), 227–234 (2007) Gutteling, J.J., et al.: Determinants of quality of life in chronic liver patients. Aliment Pharmacol Ther 23(11), 1629–1635 (2006) Unal, G., et al.: A psychometric comparison of health-related quality of life measures in chronic liver disease. J Clin Epidemiol 54(6), 587–596 (2001) Younossi, Z.M., et al.: Development of a disease specific questionnaire to measure health related quality of life in patients with chronic liver disease. Gut 45(2), 295–300 (1999) Gralnek, I.M., et al.: Development and evaluation of the Liver Disease Quality of Life instrument in persons with advanced,
Preferences for antiviral therapy of chronic hepatitis C
29. 30.
31. 32.
33. 34.
35.
36.
37.
38.
39.
40.
chronic liver disease-the LDQOL 1.0. Am J Gastroenterol 95(12), 3552–3565 (2000) Moser, G.: Hepatitis C und Psyche. J Gastroenterol Hepatol Erkrank 2(3), 20–22 (2004) Sarrazin, C., et al.: Prophylaxis, diagnosis and therapy of hepatitis C virus (HCV) infection: the German guidelines on the management of HCV infection. Z Gastroenterol 48(2), 289–351 (2010) Treadwell, J.R., Kearney, D., Davila, M.: Health profile preferences of hepatitis C patients. Dig Dis Sci 45(2), 345–350 (2000) Wong, J.B., et al.: Economic and clinical effects of evaluating rapid viral response to peginterferon alfa-2b plus ribavirin for the initial treatment of chronic hepatitis C. Am J Gastroenterol 98(11), 2354–2362 (2003) Craxi, A.: Early virologic response with pegylated interferons. Dig Liver Dis 36(Suppl 3), 340–343 (2004) Backus, L.I., et al.: Pretreatment assessment and predictors of hepatitis C virus treatment in US veterans coinfected with HIV and hepatitis C virus. J Viral Hepat 13(12), 799–810 (2006) Witkos, M., et al.: Predictors of antiviral therapy in a posttransfusion cohort of hepatitis C patients. Can J Gastroenterol 20(2), 107–111 (2006) Cotler, S.J., et al.: Patients’ values for health states associated with hepatitis C and physicians’ estimates of those values. Am J Gastroenterol 96(9), 2730–2736 (2001) Fraenkel, L., et al.: Patients’ preferences for treatment of hepatitis C. Med Decis Mak: Int J Soc Med Decis Mak 30(1), 45–57 (2010) Schackman, B.R., et al.: Quality-of-life tradeoffs for hepatitis C treatment: do patients and providers agree? Med Decis Mak 28(2), 233–242 (2008) Sawtooth Software: The CBC system for choice-based conjoint analysis. In: CBC—Technical Paper (ed.) Technical Paper Series, Sawtooth Software, Inc., Washington (2008) Ryan, M., Gerard, K.: Using discrete choice experiments to value health care programmes: current practice and future research reflections. Appl Health Econ Health Policy 2(1), 55–64 (2003)
41. Street, D., Burgess, L.: The construction of optimal stated choice experiments: theory and methods. Wiley-Interscience Series in Probability and Statistics, New Jersey (2007) 42. Johnson, R.F., et al.: Constructing experimental designs for discrete-choice experiments: report of the ISPOR conjoint analysis experimental design good research practices task force. Value In Health 16(1), 3–13 (2013) 43. Orme, B.K., Sample size issues for conjoint analysis. In: Getting Started with Conjoint Analysis, 2nd edn. pp. 57–66, Research Publishers LLC, Madison (2010) 44. Bech, M., Kjaer, T., Lauridsen, J.: Does the number of choice sets matter? Results from a web survey applying a discrete choice experiment. Health Econ 20(3), 273–286 (2011) 45. Brett Hauber, A., et al.: Patient preferences and assessment of likely adherence to hepatitis C virus treatment. J Viral Hepatitis 18(9), 619–627 (2011) 46. Kauf, T.L., et al.: Patients’ willingness to accept the risks and benefits of new treatments for chronic hepatitis C virus infection. Patient: Patient-Cent Outcomes Res 5(4), 265–278 (2012) 47. Manns, M.P., Cornberg, M.: Sofosbuvir: the final nail in the coffin for hepatitis C? Lancet Infect Dis 13(5), 378–379 (2013) 48. Schneider, M., Zeuzem, S., Sarrazin, C.: Aktueller Stand der Hepatitis-C-Therapie. Der Gastroenterologe 8(1), 27–34 (2013) 49. Koch-Institut, Robert: Zur Situation bei wichtigen Infektionskrankheiten in Deutschland: virushepatitis B, C und D im Jahr 2010. Epidemiol Bull 29, 261–274 (2010) 50. Shepard, C.W., et al.: Cost-effectiveness of conjugate meningococcal vaccination strategies in the United States. Pediatrics 115(5), 1220–1232 (2005) 51. Ryan, M., Gerard, K., Amaya-Amaya M., Using discrete choice experiments to value health and health care. In: The Economics of Non-Market Goods and Resources, 11, p. XIX 254 S, Springer, Dordrecht 2008 52. Institut fu¨r Qualita¨t und Wirtschaftlicheit im Gesundheitswesen (Institute for Quality and Efficiency in Health Care) (IQWiG), Allgemeine Methoden 4.2. Ko¨ln (2015)
123
