- Email: [email protected]
Prospective Utility Study of Patients with Multiple Cardiovascular Events
A348
VA L U E I N H E A LT H 1 9 ( 2 0 1 6 ) A 3 4 7 – A 7 6 6
HT2 A Review of the Use of Network Meta-Analysis In Nice Single Technology Appraisals Fleetwood K1, Glanville J2, McCool R2, Wood H2, Wilson K2, Marshall C2, Yellowlees A1, James D1, Toupin S1, McCabe R1 1Quantics Consulting Ltd, Edinburgh, UK, 2York Health Economics Consortium Ltd, York, UK
Objectives: The National Institute for Health and Care Excellence (NICE) evaluates the use of treatments within the English National Health Service (NHS). Within NICE’s Single Technology Appraisal (STA) process manufacturers submit evidence on the effectiveness of their interventions relative to specific comparators. The manufacturers’ submissions are evaluated by Evidence Review Groups (ERGs), and Appraisal Committees decide whether to recommend the treatment. Evidence on the relative effectiveness of the intervention and its comparators should ideally come from head-to-head randomized clinical trials. When such evidence is unavailable manufacturers may use a network meta-analysis (NMA) to compare the treatments. Both the NICE Guide to the Methods of Technology Appraisal and the NICE Decision Support Unit (DSU) Technical Support Documents (TSDs) provide guidance on the conduct of NMAs. The purpose of this review is to evaluate the use of NMA within STAs with respect to the NICE guidance. Methods: We used the NICE website to identify STAs published between May 2015 and April 2016 inclusive. For STAs that included an NMA we reviewed the appraisal documents including the manufacturer’s submission and the ERG report. Specifically we recorded what outcomes were analysed, what methods were applied, how the results were presented and any feedback from the ERG. Results: Our review demonstrated that NMAs are widely used within STAs. Manufacturers applied both standard NMA methodology as described in the NICE DSU TSDs and more complex methods such as matching adjusted indirect treatment comparison. Manufacturers did not consistently report all of the details suggested by the NICE guidelines. The ERGs differed in their expectations for NMA. In some cases, ERGs were willing to accept more complex methods. Conclusions: Although STAs often include NMAs, these do not always entirely conform to the NICE guidelines. Manufacturers should present all of the information recommended by the NICE guidelines. HT3 Actuarial Approaches To Modelling and Mitigating Financial Uncertainty in Recommending New Drugs and Health Technologies Serre D, Buckle J Milliman, London, UK
Health technology assessments (HTAs) are fundamental in informing decisions on reimbursement, assessing whether premium drug pricing is commensurate with the expected incremental health benefits. In collaboration with the National Institute for Health and Care Excellence (NICE) in England, alternative methodological approaches to the appraisals of new drugs and technologies are explored, which offer valuable perspectives on handling financial uncertainty. Building on the evidence from recent hepatitis C manufacturer submissions to NICE, a statedependent Markov model is replicated as a proof of concept to demonstrate the degree of financial uncertainty around recommending new drugs for routine commissioning. A theoretical framework is developed to illustrate how risk sharing agreements focussing on parameters and assumptions with the greatest potential budget impact can be implemented. Selected risk mitigation strategies prevalent in insurance settings such as stop-loss schemes and risk corridors support the design of one-way and two-way risk sharing agreements. These are tailored to meet the specificities of the model assumptions from the case study. Recognising the strong influence of the mean incremental cost-effectiveness ratio (ICER) in decisionmaking, but also the lack of a defined framework for appraising uncertainty around this ratio, reimbursement decisions based on the median ICER and a 75% percentile ICER are explored in the simulation. A measure of variability such as the coefficient of variation supplements the proposed approaches, which incorporate financial uncertainty more explicitly due to the allocation of probability distributions to cost and outcome metrics. Finally, the methodology reinforces the need to monitor actual experience against projections through retrospective reviews of historical medical data, to inform the value of the recoverable per the risk sharing terms. It further builds on actuarial principles and stochastic modelling for handling the financial yet uncertain impact of recommending new drugs and health technologies in a context of value-based care.
HT4 Implementation of Conditional Reimbursement Schemes in HTA Practice: Experiences from the Netherlands Makady A1, Nijmeijer H2, de Boer A3, Hillege JL4, Klungel O3, Goettsch W1 1The National Healthcare Institute (ZIN), Diemen, The Netherlands, 2Radboud University Medical Centre, Nijmegen, The Netherlands, 3Utrecht University, Utrecht, The Netherlands, 4University Medical Center Groningen, Groningen, The Netherlands
Objectives: In 2007, the National Healthcare Institute (ZIN) initiated conditional financing (CF) of expensive hospital drugs as an example of conditional reimbursement schemes (CRS). CF is a 4-year procedure encompassing initial HTA assessment (T= 0) followed by additional data collection via outcomes research (separately assessing appropriate use & cost effectiveness in routine practice) and re-assessment (T= 4). This study aims to review performance and experiences with CF in the Netherlands to date. Methods: All dossiers for drugs that underwent the full CF procedure were reviewed. Using a standardised data abstraction form, 2 researchers independently extracted information on procedural and methodological aspects (i.e. related to implemented outcomes research & evidence assessment). A scoring algorithm was used to assess both aspects. Results: Forty-seven candidates were nominated for CF; 44 underwent T= 0 assessments and 10 T= 4 assessments. The CF procedure extended beyond 4 years for 8 of the 10 candidates. For the 10 candidates, applicants clearly defined study designs and data collection methods for outcomes
research proposals addressing 14 of 20 research questions posed in T= 0 reports. ZIN provided discussion points and recommendations regarding research proposals for 16 of 20 research questions. Applicants implemented recommendations fully in 2 cases and partially in 14. Sufficient data was available at T= 4 to answer 13 of 20 research questions posed at T= 0. However, discussion points remained regarding implemented outcomes research for all 10 candidates at T= 4. ZIN advised to continue reimbursement for 6 candidates and to stop reimbursement for 2. In 2 of the 6 candidates, reimbursement was continued on the condition of additional evidence generation beyond T= 4. Two candidates did not receive a final recommendation. Conclusions: Theoretically, CF provides a valuable option for enabling quick but conditional access to medicines in the Netherlands. However, procedural and methodological aspects related to scheme design and implementation may affect its value in decision-making practice.
UTILITY STUDIES UT1 Utility By Treatment Line In Multiple Myeloma; Analysis of Over 9000 EQ-5Q-3L Questionnaires from the Emmos Registry Hatswell AJ1, Couturier C2, Ito T3 1BresMed, Sheffield, UK, 2Janssen Medical Affairs, Paris, France, 3Janssen Health Economics & Market Access EMEA, High Wycombe, UK
Objectives: Health state utilities are used in economic modelling to represent the health-related quality of life (HRQL) of patients, with the EQ-5D-3L being the most commonly used tool across disease areas. The objective of this study was to investigate the HRQL of patients with multiple myeloma (MM) across the treatment pathway in a single dataset not limited to one treatment line (as is commonly seen with trial data), and as such, produce an internally consistent set of utility values. Methods: The EMMOS registry (NCT01241396) full analysis set contains data from 2358 subjects with MM enrolled and followed at 234 sites in 22 countries from 2010 to 2014, including 9080 completed EQ-5D-3L questionnaires. These were scored using the UK algorithm, and analysed using Generalised Estimating Equation techniques. A treatment was considered a new treatment line when a class of treatment (proteasome inhibitor, immunomodulatory agent, chemotherapy) was changed. Stem Cell Transplant [SCT] was also included in the analysis. Results: Patients with newly diagnosed untreated MM have a low utility (approximately 0.46), which increases to approximately 0.60 whilst receiving their first two treatment lines, falling to approximately 0.55 beyond this. The inclusion of prior SCT as a dummy variable increases the explanatory power of the analysis. SCT is associated with a 0.13 increase in utility - even when controlling for the number of prior treatment lines. Conclusions: This analysis is the most comprehensive available in the disease area, tracking patients through multiple lines of treatment within a single study. The finding that SCT is influential in HRQL (regardless of patient characteristics) should also be factored in to future analyses. The mechanism for this finding however is not clear – it may be that SCT increases HRQL directly, or that only the healthiest patients are able to receive SCT and as such, it represents patient selection. UT2 A Meta Regression Analysis of Utility Weights for Breast Cancer Gong JR, Bae S, Lim J Ewha Womans University, Seoul, South Korea
Objectives: The purpose of meta-analysis is to derive pooled utility in different stages of breast cancer and to assess the relative impacts of study design characteristics in predicting utilities for breast cancer. Methods: We searched Medline, Embase, RISS, and KoreaMed to find out literatures reporting all publicly available utilities for breast cancer. Twenty six articles were identified reporting 124 unique utilities. We extracted 13 variables which were given in the articles: 1)mean utility weight 2) standard errors 3)assessment methods 4)disease stages 5)types of respondents, 6)number of respondents 7) countries of the respondents 8) age of the respondents 9)treatment methods 10)the lower bounds 11)the upper bounds of the scale 12)survey methods 13) survey origin. We performed a meta-regression with above variables being independent variables. Results: We found that the pooled utility was 0.6995 for the total breast cancer, 0.8254 for the early stage, 0.711 for the locally advanced stage, and 0.5569 for the recurrent and metastatic stage(P< 0.0001). Assessment method, survey origin and respondent type show significant differences in the early stage(P< 0.05), survey origin, survey method, and the lower and upper bound for the scale were significant in the recurrent and metastatic stage(P< 0.05). The results of the regression analysis revealed that the severity of breast cancer, assessment method, and survey origin were significant predictors of quality of life while respondent type, age, survey method, and the lower and upper bound of the scale were not significant. Utility weight for recurrent and metastatic stage was lower than the values which was estimated for the early stage(P< 0.0001). Utility weight estimated with scenario was lower than the results with own health status(P< 0.05). Conclusions: Quality of life estimates for breast cancer show different values in the same health status by assessment method, survey origin, survey method, and respondent type.
UT3 Prospective Utility Study of Patients with Multiple Cardiovascular Events Pockett RD1, McEwan P2, Ray J3, Tran I4, Shutler S5, Martin S6, Yousef Z7, Bakhai A8 1Swansea University, Swansea, UK, 2Health Economics and Outcomes Research Ltd, Cardiff, UK, 3F. Hoffmann-La Roche, Basel, Switzerland, 4Roche Products Ltd, Welwyn Garden City, UK, 5F. Hoffmann-La Roche Ltd., Basel, Switzerland, 6Peterborough and Stamford Hospitals NHS Foundation Trust, Peterborough, UK, 7University Hospital of Wales, Cardiff, UK, 8Royal Free London NHS Foundation Trust, Barnet, UK
VA L U E I N H E A LT H 1 9 ( 2 0 1 6 ) A 3 4 7 – A 7 6 6
Objectives: The effects of acute coronary syndrome (ACS) events on healthrelated quality of life (HRQoL) and the time dependency of these effects are unknown. The aim of this study is to characterise health utilities in ACS patients. This will help development of future economic models estimating the cost per quality adjusted life year impact of ACS events and potential treatments. Methods: Multicentre, non-interventional, longitudinal evaluation of health utility in patients experiencing ACS or stroke events. EuroQol-5 dimension surveys were sent to patients (≥ 18 years) from three centres in the UK 1 month following hospital admission for myocardial infarction (MI), unstable angina (UA) or stroke. Patient demographics, lifestyle and baseline utility score were collected in the first survey. Follow-up surveys were sent at 6, 12, 18 and 24 months to prospectively capture utility and subsequent health events. A group of patients were also identified retrospectively and patient demographics, lifestyle, and time since previous ACS event were collected. General healthy population utility values were assumed for pre-event HRQoL. Results: Between January 2011 and March 2014, 2103 prospectively/retrospectively identified patients (mean age 68.3 [range 24–97] years; 67.9% male) responded: 1176 (55.9%) MI; 898 (42.7%) UA; 29 (1.4%) stroke; 24% had type 2 diabetes. Utility values post-event were lower than general healthy population values, although significant differences in utility among subsequent 6, 18, 12 and 24-month timepoints were not detected. However, a significant difference in utility between 12 and 18 months for the retrospectively identified subgroup only was observed. Through multivariate regressions analyses, wheelchair use, current smoking and secondary mental and joint health events were associated with the greatest utility decrements (> 0.250 decrease). Conclusions: This study indicates that health utility decreases following a CV event and, while some improvement occurs over the subsequent 24 months, general healthy population utility is not necessarily attained.
UT4 Mapping Quality of Life Scores from FACT-G, FAACT And FACIT-F Onto Preference-Based Utilities Using the 5-Level Version of EQ-5D Questionnaire Meregaglia M1, Borsoi L1, Cairns J2, Tarricone R1 University, Milan, Italy, 2London School of Hygiene and Tropical Medicine, London, UK
1Bocconi
Objectives: The aim of this study was to develop and validate mapping algorithms to predict EQ-5D-5L utilities from two questionnaires (Functional Assessment of Anorexia/Cachexia Treatment – FAACT and Functional Assessment of Chronic Illness Therapy-Fatigue – FACIT-F) and their common component (Functional Assessment of Cancer Therapy-General – FACT-G) in patients with non-small cell lung cancer - cachexia (NSCLC-C). Methods: Data were collected at five occasions over a 12-week period in two multicenter, placebo-controlled trials (ROMANA 1 and ROMANA 2). The study sample was divided into development and validation datasets according to patient’s country of origin. Generalized estimating equations (GEEs) were performed to predict EQ-5D utilities from FACT-G, FAACT and FACIT-F scores. Five different sets of independent variables were tested including overall, Trial Outcome Index (TOI) and individual subscales results. The best performing models were selected based on mean absolute error (MAE) and root-mean square error (RMSE). Results: A subset of 96 patients completed both EQ-5D-5L and FAACT/FACIT-F questionnaires. Models using the individual domains separately yielded the lowest MAE/RMSE in most of study time points; however, even algorithms modeling the overall scores showed a high predictive performance. In FACT-G models, Physical Well-Being had the highest explanatory value (0.0094; p< 0.001), while Emotional Well-Being did not significantly affect the EQ-5D score; AnorexiaCachexia (0.0035; p= 0.007) and Fatigue (0.0059; p< 0.001) subscales were highly statistically significant in FAACT and FACIT-F models, respectively. The Eastern Cooperative Oncology Group status was the only covariate retained in the final models after backward selection. All the differences between mean observed and predicted EQ-5D utility were below the Minimal Important Difference (0.08) established in cancer for UK-index scores. Conclusions: The developed algorithms enable the estimation of Quality-Adjusted Life Years (QALYs) from three cancer-specific instruments in cost-effectiveness analyses where EQ-5D data are missing. Further research evaluating model performance in an independent sample of NSCLC-C patients is encouraged.
BREAKOUTS – SESSION III
CARDIOVASCULAR OUTCOMES RESEARCH STUDIES CV1 Comparison of Oral Anti-Coagulants for Stroke Prevention in NonValvular Atrial Fibrillation: Two Multi-Criteria Decision Analyses Tervonen T1, Ustyugova AV2, Lip G3, Verdecchia P4, Sri Bhashyam S1, Heinrich-Nols J5, Gropper S5, Kwan R6, Marsh K1 1Evidera Ltd, London, UK, 2Boehringer Ingelheim GmbH, Ingelheim am Rhein, Germany, 3University of Birmingham, Birmingham, UK, 4Hospital of Assisi, Assisi, Italy, 5Boehringer Ingelheim Pharma GmbH & Co. KG, Ingelheim am Rhein, Germany, 6Boehringer Ingelheim (Canada) Ltd, Burlington, ON, Canada
Objectives: To compare five oral anticoagulants (OACs) available in the UK for stroke prevention in patients suffering from non-valvular atrial fibrillation (NVAF), based on factors relevant for payers and prescribers. Methods: Two multi-criteria decision analyses (MCDA) were developed to compare apixaban, dabigatran, edoxaban, rivaroxaban and a vitamin K antagonist (VKA; i.e., warfarin) from payer and prescriber perspectives. Final evaluation models included up to ten clinical and three non-clinical criteria. The clinical criteria were ranked based on either expected changes in event mortality rates from the worst to best
A349
performing treatment, as demonstrated in clinical trials, or based on variation of health-related costs (both acute and follow-up for one year). These rankings were used to compute centroid weights. An additive model was used to combine treatment performance with centroid weights to estimate the overall value of each OAC. Probabilistic and structural sensitivity analyses were conducted. Results: Dabigatran was the best treatment in centroid weight analyses with 7% / 8% higher overall value than the second best performing treatment, apixaban. Dabigatran also had the highest first rank probability (72% / 70%) in probabilistic sensitivity analyses, with apixaban being second (22% / 24% first rank probability). Rivaroxaban performed worse than other non-VKA OACs, but better than VKA (both with 0% first rank probability). The results were largely insensitive to changes in model structure, although changing availability of reversal agent to be the least important criterion increased apixaban to have approximately the same overall value as dabigatran. Conclusions: Despite using only rank-based preference data, we were able to demonstrate dabigatran to be the most and warfarin the least preferred treatment with an MCDA incorporating all factors relevant for distinguishing OACs. CV2 Using Subpopulation Treatment Effect Pattern Plot to Identify More Efficient Resource Allocation Policies Cao Q1, Hillege JL2, Postma MJ2, Buskens E2, Postmus D3 1University of Groningen, Groningen, The Netherlands, 2University Medical Center Groningen, Groningen, The Netherlands, 3University of Groningen, University Medical Center Groningen, Groningen, The Netherlands
Objectives: When cost-effectiveness analyses are conducted alongside randomized controlled trials it is important to acknowledge patient heterogeneity as this may result in more efficient resource allocation policies. In this study, we sought to explore to what extent the use of Subpopulation Treatment Effect Pattern Plot (STEPP) may facilitate such subgroup analysis strategies. Methods: The analysis was based on data from the COACH study, in which 1,023 patients with heart failure were randomly assigned to three treatments: care-as-usual, basic support, and intensive support. First, using predicted 18-month mortality risk as the stratification basis, a suitable strategy for assigning different treatments to different risk groups of patient was developed. To that end a graphical exploration of the difference in net monetary benefit (NMB) across treatment regimens and baseline risk was used. Next, the efficiency gains resulting from this proposed subgroup strategy were quantified by computing the difference in NMB between our stratified approach and the best performing population-wide strategy. Results: The STEPP approach allowed distinguishing between subgroups, i.e., intensive support appeared optimal for low-risk patients (18-month mortality risk ≤ 0.16), while basic support appeared optimal for intermediate to high-risk patients (18-month mortality risk > 0.16). The average gain in NMB resulting from a stratified approach compared to basic support for all was € 1,312 (95% CI: € 390-€ 2,346). Conclusions: A risk-based analysis using STEPP seems promising to explore the impact of baseline risk for the relative cost-effectiveness in optimizing treatment trade-off and subsequently in the quest for more efficient reimbursement policies. CV3 Policy Objective of Generic Medicines from the Investment Perspective: The Case of Clopidogrel Elek P1, Harsányi A2, Zelei T3, Csetneki K3, Kaló Z3 1Eötvös Loránd University (ELTE), Budapest, Hungary, 2Eötvös Loránd University, Budapest, Hungary, 3Syreon Research Institute, Budapest, Hungary
Objectives: The objective of generic drug policies in most countries is defined from a disinvestment perspective: reduction in expenditures without compromising health outcomes. However, in countries with restricted access of patients to original patented drugs, the objective of generic drug policies can also be defined from an investment perspective: health gain by improved patient access without need for additional health budget. The objective of this study was to assess whether generic clopidogrel entry reduced the role of affordability constraints and increased clopidogrel utilization in European countries. Methods: We analyzed the determinants of clopidogrel utilization in Europe between 2004 and 2014 using hierarchical linear models on country-level longitudinal data. The first generic clopidogrel entry occurred in 2009 in the majority of countries. Results: Clopidogrel utilization was strongly affected by affordability constraints (as proxied by GDP per capita) before entrance of generic medicines, but this effect decayed by 2014. Our estimated hierarchical linear models found a substantially larger trend increase of clopidogrel utilization in lower-income European countries than in the higher-income ones. Similarly, generic entry increased clopidogrel consumption in lower-income countries but did not have an effect in the highest-income ones. The models also suggest that an earlier generic entry was associated with a larger effect. Conclusions: The case of clopidogrel indicates that the entrance of generics may increase patient access to effective medicines, most notably in lower-income countries, thereby reducing inequalities between European patients. Policymakers should also consider this investment aspect of generic medicines when designing international and national pharmaceutical policies. CV4 Are Component Endpoints Equal? A Study into the Practice of Composite End Points in Clinical Trials Vaanholt MC1, von Birgelen C2, Kok M3, Weernink MG1, van Til J2 Netherlands, 2University of Twente, Enschede, The Netherlands, 3Medisch Spectrum Twente, Enschede, The Netherlands
1University Twente, Enschede, The
Objectives: Clinical trials comparing treatments for coronary revascularization generally use composite end points in order to increase statistical precision and
VA L U E I N H E A LT H 1 9 ( 2 0 1 6 ) A 3 4 7 – A 7 6 6
HT2 A Review of the Use of Network Meta-Analysis In Nice Single Technology Appraisals Fleetwood K1, Glanville J2, McCool R2, Wood H2, Wilson K2, Marshall C2, Yellowlees A1, James D1, Toupin S1, McCabe R1 1Quantics Consulting Ltd, Edinburgh, UK, 2York Health Economics Consortium Ltd, York, UK
Objectives: The National Institute for Health and Care Excellence (NICE) evaluates the use of treatments within the English National Health Service (NHS). Within NICE’s Single Technology Appraisal (STA) process manufacturers submit evidence on the effectiveness of their interventions relative to specific comparators. The manufacturers’ submissions are evaluated by Evidence Review Groups (ERGs), and Appraisal Committees decide whether to recommend the treatment. Evidence on the relative effectiveness of the intervention and its comparators should ideally come from head-to-head randomized clinical trials. When such evidence is unavailable manufacturers may use a network meta-analysis (NMA) to compare the treatments. Both the NICE Guide to the Methods of Technology Appraisal and the NICE Decision Support Unit (DSU) Technical Support Documents (TSDs) provide guidance on the conduct of NMAs. The purpose of this review is to evaluate the use of NMA within STAs with respect to the NICE guidance. Methods: We used the NICE website to identify STAs published between May 2015 and April 2016 inclusive. For STAs that included an NMA we reviewed the appraisal documents including the manufacturer’s submission and the ERG report. Specifically we recorded what outcomes were analysed, what methods were applied, how the results were presented and any feedback from the ERG. Results: Our review demonstrated that NMAs are widely used within STAs. Manufacturers applied both standard NMA methodology as described in the NICE DSU TSDs and more complex methods such as matching adjusted indirect treatment comparison. Manufacturers did not consistently report all of the details suggested by the NICE guidelines. The ERGs differed in their expectations for NMA. In some cases, ERGs were willing to accept more complex methods. Conclusions: Although STAs often include NMAs, these do not always entirely conform to the NICE guidelines. Manufacturers should present all of the information recommended by the NICE guidelines. HT3 Actuarial Approaches To Modelling and Mitigating Financial Uncertainty in Recommending New Drugs and Health Technologies Serre D, Buckle J Milliman, London, UK
Health technology assessments (HTAs) are fundamental in informing decisions on reimbursement, assessing whether premium drug pricing is commensurate with the expected incremental health benefits. In collaboration with the National Institute for Health and Care Excellence (NICE) in England, alternative methodological approaches to the appraisals of new drugs and technologies are explored, which offer valuable perspectives on handling financial uncertainty. Building on the evidence from recent hepatitis C manufacturer submissions to NICE, a statedependent Markov model is replicated as a proof of concept to demonstrate the degree of financial uncertainty around recommending new drugs for routine commissioning. A theoretical framework is developed to illustrate how risk sharing agreements focussing on parameters and assumptions with the greatest potential budget impact can be implemented. Selected risk mitigation strategies prevalent in insurance settings such as stop-loss schemes and risk corridors support the design of one-way and two-way risk sharing agreements. These are tailored to meet the specificities of the model assumptions from the case study. Recognising the strong influence of the mean incremental cost-effectiveness ratio (ICER) in decisionmaking, but also the lack of a defined framework for appraising uncertainty around this ratio, reimbursement decisions based on the median ICER and a 75% percentile ICER are explored in the simulation. A measure of variability such as the coefficient of variation supplements the proposed approaches, which incorporate financial uncertainty more explicitly due to the allocation of probability distributions to cost and outcome metrics. Finally, the methodology reinforces the need to monitor actual experience against projections through retrospective reviews of historical medical data, to inform the value of the recoverable per the risk sharing terms. It further builds on actuarial principles and stochastic modelling for handling the financial yet uncertain impact of recommending new drugs and health technologies in a context of value-based care.
HT4 Implementation of Conditional Reimbursement Schemes in HTA Practice: Experiences from the Netherlands Makady A1, Nijmeijer H2, de Boer A3, Hillege JL4, Klungel O3, Goettsch W1 1The National Healthcare Institute (ZIN), Diemen, The Netherlands, 2Radboud University Medical Centre, Nijmegen, The Netherlands, 3Utrecht University, Utrecht, The Netherlands, 4University Medical Center Groningen, Groningen, The Netherlands
Objectives: In 2007, the National Healthcare Institute (ZIN) initiated conditional financing (CF) of expensive hospital drugs as an example of conditional reimbursement schemes (CRS). CF is a 4-year procedure encompassing initial HTA assessment (T= 0) followed by additional data collection via outcomes research (separately assessing appropriate use & cost effectiveness in routine practice) and re-assessment (T= 4). This study aims to review performance and experiences with CF in the Netherlands to date. Methods: All dossiers for drugs that underwent the full CF procedure were reviewed. Using a standardised data abstraction form, 2 researchers independently extracted information on procedural and methodological aspects (i.e. related to implemented outcomes research & evidence assessment). A scoring algorithm was used to assess both aspects. Results: Forty-seven candidates were nominated for CF; 44 underwent T= 0 assessments and 10 T= 4 assessments. The CF procedure extended beyond 4 years for 8 of the 10 candidates. For the 10 candidates, applicants clearly defined study designs and data collection methods for outcomes
research proposals addressing 14 of 20 research questions posed in T= 0 reports. ZIN provided discussion points and recommendations regarding research proposals for 16 of 20 research questions. Applicants implemented recommendations fully in 2 cases and partially in 14. Sufficient data was available at T= 4 to answer 13 of 20 research questions posed at T= 0. However, discussion points remained regarding implemented outcomes research for all 10 candidates at T= 4. ZIN advised to continue reimbursement for 6 candidates and to stop reimbursement for 2. In 2 of the 6 candidates, reimbursement was continued on the condition of additional evidence generation beyond T= 4. Two candidates did not receive a final recommendation. Conclusions: Theoretically, CF provides a valuable option for enabling quick but conditional access to medicines in the Netherlands. However, procedural and methodological aspects related to scheme design and implementation may affect its value in decision-making practice.
UTILITY STUDIES UT1 Utility By Treatment Line In Multiple Myeloma; Analysis of Over 9000 EQ-5Q-3L Questionnaires from the Emmos Registry Hatswell AJ1, Couturier C2, Ito T3 1BresMed, Sheffield, UK, 2Janssen Medical Affairs, Paris, France, 3Janssen Health Economics & Market Access EMEA, High Wycombe, UK
Objectives: Health state utilities are used in economic modelling to represent the health-related quality of life (HRQL) of patients, with the EQ-5D-3L being the most commonly used tool across disease areas. The objective of this study was to investigate the HRQL of patients with multiple myeloma (MM) across the treatment pathway in a single dataset not limited to one treatment line (as is commonly seen with trial data), and as such, produce an internally consistent set of utility values. Methods: The EMMOS registry (NCT01241396) full analysis set contains data from 2358 subjects with MM enrolled and followed at 234 sites in 22 countries from 2010 to 2014, including 9080 completed EQ-5D-3L questionnaires. These were scored using the UK algorithm, and analysed using Generalised Estimating Equation techniques. A treatment was considered a new treatment line when a class of treatment (proteasome inhibitor, immunomodulatory agent, chemotherapy) was changed. Stem Cell Transplant [SCT] was also included in the analysis. Results: Patients with newly diagnosed untreated MM have a low utility (approximately 0.46), which increases to approximately 0.60 whilst receiving their first two treatment lines, falling to approximately 0.55 beyond this. The inclusion of prior SCT as a dummy variable increases the explanatory power of the analysis. SCT is associated with a 0.13 increase in utility - even when controlling for the number of prior treatment lines. Conclusions: This analysis is the most comprehensive available in the disease area, tracking patients through multiple lines of treatment within a single study. The finding that SCT is influential in HRQL (regardless of patient characteristics) should also be factored in to future analyses. The mechanism for this finding however is not clear – it may be that SCT increases HRQL directly, or that only the healthiest patients are able to receive SCT and as such, it represents patient selection. UT2 A Meta Regression Analysis of Utility Weights for Breast Cancer Gong JR, Bae S, Lim J Ewha Womans University, Seoul, South Korea
Objectives: The purpose of meta-analysis is to derive pooled utility in different stages of breast cancer and to assess the relative impacts of study design characteristics in predicting utilities for breast cancer. Methods: We searched Medline, Embase, RISS, and KoreaMed to find out literatures reporting all publicly available utilities for breast cancer. Twenty six articles were identified reporting 124 unique utilities. We extracted 13 variables which were given in the articles: 1)mean utility weight 2) standard errors 3)assessment methods 4)disease stages 5)types of respondents, 6)number of respondents 7) countries of the respondents 8) age of the respondents 9)treatment methods 10)the lower bounds 11)the upper bounds of the scale 12)survey methods 13) survey origin. We performed a meta-regression with above variables being independent variables. Results: We found that the pooled utility was 0.6995 for the total breast cancer, 0.8254 for the early stage, 0.711 for the locally advanced stage, and 0.5569 for the recurrent and metastatic stage(P< 0.0001). Assessment method, survey origin and respondent type show significant differences in the early stage(P< 0.05), survey origin, survey method, and the lower and upper bound for the scale were significant in the recurrent and metastatic stage(P< 0.05). The results of the regression analysis revealed that the severity of breast cancer, assessment method, and survey origin were significant predictors of quality of life while respondent type, age, survey method, and the lower and upper bound of the scale were not significant. Utility weight for recurrent and metastatic stage was lower than the values which was estimated for the early stage(P< 0.0001). Utility weight estimated with scenario was lower than the results with own health status(P< 0.05). Conclusions: Quality of life estimates for breast cancer show different values in the same health status by assessment method, survey origin, survey method, and respondent type.
UT3 Prospective Utility Study of Patients with Multiple Cardiovascular Events Pockett RD1, McEwan P2, Ray J3, Tran I4, Shutler S5, Martin S6, Yousef Z7, Bakhai A8 1Swansea University, Swansea, UK, 2Health Economics and Outcomes Research Ltd, Cardiff, UK, 3F. Hoffmann-La Roche, Basel, Switzerland, 4Roche Products Ltd, Welwyn Garden City, UK, 5F. Hoffmann-La Roche Ltd., Basel, Switzerland, 6Peterborough and Stamford Hospitals NHS Foundation Trust, Peterborough, UK, 7University Hospital of Wales, Cardiff, UK, 8Royal Free London NHS Foundation Trust, Barnet, UK
VA L U E I N H E A LT H 1 9 ( 2 0 1 6 ) A 3 4 7 – A 7 6 6
Objectives: The effects of acute coronary syndrome (ACS) events on healthrelated quality of life (HRQoL) and the time dependency of these effects are unknown. The aim of this study is to characterise health utilities in ACS patients. This will help development of future economic models estimating the cost per quality adjusted life year impact of ACS events and potential treatments. Methods: Multicentre, non-interventional, longitudinal evaluation of health utility in patients experiencing ACS or stroke events. EuroQol-5 dimension surveys were sent to patients (≥ 18 years) from three centres in the UK 1 month following hospital admission for myocardial infarction (MI), unstable angina (UA) or stroke. Patient demographics, lifestyle and baseline utility score were collected in the first survey. Follow-up surveys were sent at 6, 12, 18 and 24 months to prospectively capture utility and subsequent health events. A group of patients were also identified retrospectively and patient demographics, lifestyle, and time since previous ACS event were collected. General healthy population utility values were assumed for pre-event HRQoL. Results: Between January 2011 and March 2014, 2103 prospectively/retrospectively identified patients (mean age 68.3 [range 24–97] years; 67.9% male) responded: 1176 (55.9%) MI; 898 (42.7%) UA; 29 (1.4%) stroke; 24% had type 2 diabetes. Utility values post-event were lower than general healthy population values, although significant differences in utility among subsequent 6, 18, 12 and 24-month timepoints were not detected. However, a significant difference in utility between 12 and 18 months for the retrospectively identified subgroup only was observed. Through multivariate regressions analyses, wheelchair use, current smoking and secondary mental and joint health events were associated with the greatest utility decrements (> 0.250 decrease). Conclusions: This study indicates that health utility decreases following a CV event and, while some improvement occurs over the subsequent 24 months, general healthy population utility is not necessarily attained.
UT4 Mapping Quality of Life Scores from FACT-G, FAACT And FACIT-F Onto Preference-Based Utilities Using the 5-Level Version of EQ-5D Questionnaire Meregaglia M1, Borsoi L1, Cairns J2, Tarricone R1 University, Milan, Italy, 2London School of Hygiene and Tropical Medicine, London, UK
1Bocconi
Objectives: The aim of this study was to develop and validate mapping algorithms to predict EQ-5D-5L utilities from two questionnaires (Functional Assessment of Anorexia/Cachexia Treatment – FAACT and Functional Assessment of Chronic Illness Therapy-Fatigue – FACIT-F) and their common component (Functional Assessment of Cancer Therapy-General – FACT-G) in patients with non-small cell lung cancer - cachexia (NSCLC-C). Methods: Data were collected at five occasions over a 12-week period in two multicenter, placebo-controlled trials (ROMANA 1 and ROMANA 2). The study sample was divided into development and validation datasets according to patient’s country of origin. Generalized estimating equations (GEEs) were performed to predict EQ-5D utilities from FACT-G, FAACT and FACIT-F scores. Five different sets of independent variables were tested including overall, Trial Outcome Index (TOI) and individual subscales results. The best performing models were selected based on mean absolute error (MAE) and root-mean square error (RMSE). Results: A subset of 96 patients completed both EQ-5D-5L and FAACT/FACIT-F questionnaires. Models using the individual domains separately yielded the lowest MAE/RMSE in most of study time points; however, even algorithms modeling the overall scores showed a high predictive performance. In FACT-G models, Physical Well-Being had the highest explanatory value (0.0094; p< 0.001), while Emotional Well-Being did not significantly affect the EQ-5D score; AnorexiaCachexia (0.0035; p= 0.007) and Fatigue (0.0059; p< 0.001) subscales were highly statistically significant in FAACT and FACIT-F models, respectively. The Eastern Cooperative Oncology Group status was the only covariate retained in the final models after backward selection. All the differences between mean observed and predicted EQ-5D utility were below the Minimal Important Difference (0.08) established in cancer for UK-index scores. Conclusions: The developed algorithms enable the estimation of Quality-Adjusted Life Years (QALYs) from three cancer-specific instruments in cost-effectiveness analyses where EQ-5D data are missing. Further research evaluating model performance in an independent sample of NSCLC-C patients is encouraged.
BREAKOUTS – SESSION III
CARDIOVASCULAR OUTCOMES RESEARCH STUDIES CV1 Comparison of Oral Anti-Coagulants for Stroke Prevention in NonValvular Atrial Fibrillation: Two Multi-Criteria Decision Analyses Tervonen T1, Ustyugova AV2, Lip G3, Verdecchia P4, Sri Bhashyam S1, Heinrich-Nols J5, Gropper S5, Kwan R6, Marsh K1 1Evidera Ltd, London, UK, 2Boehringer Ingelheim GmbH, Ingelheim am Rhein, Germany, 3University of Birmingham, Birmingham, UK, 4Hospital of Assisi, Assisi, Italy, 5Boehringer Ingelheim Pharma GmbH & Co. KG, Ingelheim am Rhein, Germany, 6Boehringer Ingelheim (Canada) Ltd, Burlington, ON, Canada
Objectives: To compare five oral anticoagulants (OACs) available in the UK for stroke prevention in patients suffering from non-valvular atrial fibrillation (NVAF), based on factors relevant for payers and prescribers. Methods: Two multi-criteria decision analyses (MCDA) were developed to compare apixaban, dabigatran, edoxaban, rivaroxaban and a vitamin K antagonist (VKA; i.e., warfarin) from payer and prescriber perspectives. Final evaluation models included up to ten clinical and three non-clinical criteria. The clinical criteria were ranked based on either expected changes in event mortality rates from the worst to best
A349
performing treatment, as demonstrated in clinical trials, or based on variation of health-related costs (both acute and follow-up for one year). These rankings were used to compute centroid weights. An additive model was used to combine treatment performance with centroid weights to estimate the overall value of each OAC. Probabilistic and structural sensitivity analyses were conducted. Results: Dabigatran was the best treatment in centroid weight analyses with 7% / 8% higher overall value than the second best performing treatment, apixaban. Dabigatran also had the highest first rank probability (72% / 70%) in probabilistic sensitivity analyses, with apixaban being second (22% / 24% first rank probability). Rivaroxaban performed worse than other non-VKA OACs, but better than VKA (both with 0% first rank probability). The results were largely insensitive to changes in model structure, although changing availability of reversal agent to be the least important criterion increased apixaban to have approximately the same overall value as dabigatran. Conclusions: Despite using only rank-based preference data, we were able to demonstrate dabigatran to be the most and warfarin the least preferred treatment with an MCDA incorporating all factors relevant for distinguishing OACs. CV2 Using Subpopulation Treatment Effect Pattern Plot to Identify More Efficient Resource Allocation Policies Cao Q1, Hillege JL2, Postma MJ2, Buskens E2, Postmus D3 1University of Groningen, Groningen, The Netherlands, 2University Medical Center Groningen, Groningen, The Netherlands, 3University of Groningen, University Medical Center Groningen, Groningen, The Netherlands
Objectives: When cost-effectiveness analyses are conducted alongside randomized controlled trials it is important to acknowledge patient heterogeneity as this may result in more efficient resource allocation policies. In this study, we sought to explore to what extent the use of Subpopulation Treatment Effect Pattern Plot (STEPP) may facilitate such subgroup analysis strategies. Methods: The analysis was based on data from the COACH study, in which 1,023 patients with heart failure were randomly assigned to three treatments: care-as-usual, basic support, and intensive support. First, using predicted 18-month mortality risk as the stratification basis, a suitable strategy for assigning different treatments to different risk groups of patient was developed. To that end a graphical exploration of the difference in net monetary benefit (NMB) across treatment regimens and baseline risk was used. Next, the efficiency gains resulting from this proposed subgroup strategy were quantified by computing the difference in NMB between our stratified approach and the best performing population-wide strategy. Results: The STEPP approach allowed distinguishing between subgroups, i.e., intensive support appeared optimal for low-risk patients (18-month mortality risk ≤ 0.16), while basic support appeared optimal for intermediate to high-risk patients (18-month mortality risk > 0.16). The average gain in NMB resulting from a stratified approach compared to basic support for all was € 1,312 (95% CI: € 390-€ 2,346). Conclusions: A risk-based analysis using STEPP seems promising to explore the impact of baseline risk for the relative cost-effectiveness in optimizing treatment trade-off and subsequently in the quest for more efficient reimbursement policies. CV3 Policy Objective of Generic Medicines from the Investment Perspective: The Case of Clopidogrel Elek P1, Harsányi A2, Zelei T3, Csetneki K3, Kaló Z3 1Eötvös Loránd University (ELTE), Budapest, Hungary, 2Eötvös Loránd University, Budapest, Hungary, 3Syreon Research Institute, Budapest, Hungary
Objectives: The objective of generic drug policies in most countries is defined from a disinvestment perspective: reduction in expenditures without compromising health outcomes. However, in countries with restricted access of patients to original patented drugs, the objective of generic drug policies can also be defined from an investment perspective: health gain by improved patient access without need for additional health budget. The objective of this study was to assess whether generic clopidogrel entry reduced the role of affordability constraints and increased clopidogrel utilization in European countries. Methods: We analyzed the determinants of clopidogrel utilization in Europe between 2004 and 2014 using hierarchical linear models on country-level longitudinal data. The first generic clopidogrel entry occurred in 2009 in the majority of countries. Results: Clopidogrel utilization was strongly affected by affordability constraints (as proxied by GDP per capita) before entrance of generic medicines, but this effect decayed by 2014. Our estimated hierarchical linear models found a substantially larger trend increase of clopidogrel utilization in lower-income European countries than in the higher-income ones. Similarly, generic entry increased clopidogrel consumption in lower-income countries but did not have an effect in the highest-income ones. The models also suggest that an earlier generic entry was associated with a larger effect. Conclusions: The case of clopidogrel indicates that the entrance of generics may increase patient access to effective medicines, most notably in lower-income countries, thereby reducing inequalities between European patients. Policymakers should also consider this investment aspect of generic medicines when designing international and national pharmaceutical policies. CV4 Are Component Endpoints Equal? A Study into the Practice of Composite End Points in Clinical Trials Vaanholt MC1, von Birgelen C2, Kok M3, Weernink MG1, van Til J2 Netherlands, 2University of Twente, Enschede, The Netherlands, 3Medisch Spectrum Twente, Enschede, The Netherlands
1University Twente, Enschede, The
Objectives: Clinical trials comparing treatments for coronary revascularization generally use composite end points in order to increase statistical precision and