VA L U E I N H E A LT H 1 7 ( 2 0 1 4 ) A 3 2 3 – A 6 8 6 A561

losses, and informal care. Specific parameters to be entered by users are the preva-lence of the mutation, treatment costs, specificity/sensitivity and cost of the test, survival data and the incidence of AEs. ConClusions: The proposed global model for the economic evaluation of targeted treatments using companion diagnostics in advanced/metastatic cancer treatment can, with minimal input, quickly generate cost-effectiveness analyses of targeted cancer treatment.

PRM104A Web bAsed OPtiMisAtiOn MOdel FOR A PORtFOliO OF PReventAtive inteRventiOns Utilizing MUlti CRiteRiA deCisiOn AnAlyses (MCdA) FRAMeWORkTopachevskyi O.

1, Volovyk A.

2

1Digital Health Outcomes, Brussels, Belgium, 2Hashtago, Kiev, UkraineobjeCtives: To inform decision makers who seek extension of Universal Mass Vaccination (UMV) about the most optimal allocation of funds across multiple pre-ventative interventions or vaccines. To account for decision makers preferences using MCDA. Methods: A multi cohort markov model was developed to assess clinical and economic consequences of vaccine preventable diseases in Japan. Disease incidence rates, direct medical costs and QoL data were obtained from local sources. Payer perspective only was considered. Optimization module utilizing linear programming was developed to maximize outcome of interest which serve as an objective function subject to budget and intervention coverage constraints. A working version of the model can be found at http: //www. digitalho. com/models/a/portfolio/index. html. The model was initially developed in Excel and then automatically transformed into a JavaScript application to allow for an online access. One way sensitivity analyses was conducted to parametric unceranity. Results: Model results indicate that the optimal mix of interventions depends primarily on the objective function. Various single objective functions or a combination of multiple weighted objectives lead to different mix of interventions. When prevention of death is as an objective function then pneumococcal and rotavirus vaccines are chosen. ConClusions: The proposed web based model is a complementary addition to the conventional cost-effectiveness assessment for preventative interventions. This model helps to understand sequence of introduction of prevantative interventions and expected health and economic outcomes over time. The use of MCDA framework helps users to define specfic health objectives to be used in optimisation module. The web based modeling solution pro-vides a widespread access to an easy to use tool that can by used by authorities, academia and non-modeling professionals.

PRM105CAlibRAtiOn And stAtistiCAl MOdeling tO inFORM A MiCRO-siMUlAtiOn MOdel FOR eARly HtABongers M.L.

1, De Ruysscher D.

2, Oberije C.

3, Lambin P.

2, Uyl-de Groot C.A.

4, Coupe V.M.

1

1VU University Medical Center, Amsterdam, The Netherlands, 2University Hospitals Leuven/KU Leuven, Leuven, Belgium, 3MAASTRO Clinic, Maastricht, The Netherlands, 4Erasmus University Rotterdam, Rotterdam, The NetherlandsobjeCtives: For the evaluation of the potential cost-effectiveness of an early experimental therapy, we calibrated an existing micro-simulation model for radio-therapy planning in lung cancer using pilot data. Methods: We used an exter-nally validated micro-simulation model, build using Real World Evidence data. The model contained four clinical states from alive to death, with intermediate states ‘local recurrence’ and ‘metastasis’, with 5 transitions. Based on individual and time-dependent hazard rates, patients move through the model according to their combination of patient characteristics and random variation. For the experi-mental dosis-escalation therapy we had limited pilot study data, which included overall survival and a number of baseline characteristics. The distribution of patient features in the cohort of the micro-simulation model was adjusted so that the simu-lated patients had the same baseline characteristics as the patients that received experimental therapy. Alternative radiotherapy strategies affected 5 transitions in the model, quantified by 5 hazard ratios (HRs). Subsequently, HRs for experimental radiotherapy compared to current radiotherapy were calibrated until they were able to satisfactorily reproduce the survival curve of the pilot data. The best fitting sets of HRs were selected based on the least Sum of Squared Errors (SSE) of the model predictions and the survival curve of the experimental therapy on three time points. Results: The best fitting set HRs resulted in a SSE of 0,005 based on prediction errors at 1,2 and 3-year survival. Although 33 out of 1000 sets produced predictions with less than 5% prediction error, hazard ratios varied strongly within and over the different sets. ConClusions: By using calibration, we obtained a micro-simulation model that is suitable for the evaluation of new treatments in the absence of empirical data. The model will be used for cost-effectiveness analy-ses, where the variation in hazard ratios within sets will be evaluated in scenario analyses.

PRM106HOW dOes UnCeRtAinty AROUnd COsts And eFFeCts RelAte tO UnCeRtAinty AROUnd COst-eFFeCtiveness?Jain M.

1, Bhattacharyya S.

1, Gupta S.

1, Sonathi V.

1, Mahon R.

2, Malakar H.

1, Vudumala U.

1, Gunda P.

1, Kumar P.

1, Partha G.

1, Thomas S.K.

3

1Novartis Healthcare Pvt. Ltd., Hyderabad, India, 2Novartis Ireland Limited, Dublin, Ireland, 3Novartis Pharmaceuticals Corporation, East Hanover, NJ, USAobjeCtives: HTAs require information on costs and outcomes as well as the uncertainty around them for making reimbursement decisions. Uncertainty around costs and effects (outcome uncertainty) can be substantial and increas-ingly so at more distal time points. However, the uncertainty surrounding the decision to adopt or reject a technology based on cost-effectiveness (decision uncertainty) evolves over time in a different manner. In this analysis, we intend to illustrate that increased outcome uncertainty need not result in increased decision uncertainty and that both may evolve over time differently. Methods: A previously published lifetime Markov model, built from UK health care perspec-tive, was used in the analysis. The model compared the cost-effectiveness (CE) of

PRM101APPliCAtiOn OF A MOdel OF deCisiOn bAsed On FUzzy lOgiC tO PHARMACOeCOnOMiCs: RAnibizUMAb veRsUs AFlibeRCeRt in AMdAlonso Herreros J.M.

1, González-Cuello A.

2

1HOSPITAL LOS ARCOS MAR MENOR, SAN JAVIER (MURCIA), Spain, 2Murcia University, MURCIA, SpainobjeCtives: The term “fuzzy logic” was introduced in 1965 by LAZadeh. Compared to traditional logic, fuzzy logic variables may have a truth value in degree. Fuzzy logic has been applied to many fields, from economic analysis, to artificial intel-ligence. However it has not been applied so far to pharmacoeconomics. We present a model of pharmacoeconomic decision based on fuzzy logic (Fuzzy Economic Review 2001; 6 (2): 51-73) and applied to the selection of ranibizumab-aflibercept in treating AMD. Methods: According to a decision analysis model based on fuzzy logic four fuzzy variables that affect the choice of treatment are defined: treatment success (expressed as a probability), cost of success, cost of failure (expressed as inverses), and other conditions about the cost (negotiation, handling of drugs...). Based on the value of these fuzzy variables, three linguistic variables (High, Medium, Low) are defined to expressing convenience of choice. The combination of the three pos-sible values for each of the variables gives us 81 possible decision rules, so that the (HHHH) would be the most favorable option and (LLLL) the more unfavorable. So a new fuzzy variable called “ranking” is established for classifying these options with 7 possible values (Very-unfavorable, unfavorable, slightly-unfavorable, neutral, slightly-favorable, favorable, very-favorable). The value of the fuzzy variables for ranibizumab and aflibercept were established based on pivotal clinical trials at 52 weeks cited by the EMEA. Results: The matrices obtained for ranibizumab was (0.29,3. 55 10-4, -1.36 10-4 0.7), and aflibercept (0.269,7. 4 10-4 -2.59 10-4 0.3). These matrices correspond to decision rules (HLLM) and (HMML) and correspond to a ranking of “neutral” and “slightly-favorable”. ConClusions: It possible to apply methods of “fuzzy logic” to pharmacoeconomic studies to select the most favorable treatment. According to model, AMD treatment, with aflibercept would be a slightly more favorable option than ranibizumab.

PRM102develOPMent OF An inFlUenzA OUtbReAk FOReCAsting MOdel Using tiMe seRies AnAlysis MetHOdsSmolen H.J.

Medical Decision Modeling Inc., Indianapolis, IN, USAobjeCtives: To use historical influenza incidence time series data to develop a predictive model using time series analysis methods to forecast expected num-ber of reported influenza cases. bACKGRound: Influenza is a common disease associated with high mortality. Low vaccination rates motivate health officials to predict outbreaks and intervene accordingly. A predictive model would facilitate in deciding whether an apparent excess of cases represents an outbreak or a random variation. Methods: Google Flu Trend project data from 2003 to 2014 was used to construct this predictive model. The influenza time series data clearly had a sea-sonal variation to it so a seasonally fit model using seasonal indicators, a seasonally fit model using trigonometric functions, and a multiplicative seasonal autoregres-sive integrated moving average (SARIMA) model were considered. Fifty-two weeks of data from the time series were withheld from the model fitting process so as to evaluate the predictive capability of the selected model using mean absolute percentage error (MAPE). The Akaike’s Information Criterion (AIC) goodness of fit measure was used to select the model that fit the data the best (lower the bet-ter). Results: The SARIMA model provided the best fit for the data with an AIC of 6361.7. The seasonally fit model using seasonal indicators had an AIC of 8473.9 and the seasonally fit model using trigonometric functions had an AIC of 8438.2. The SARIMA model MAPE for the predicted 52 weeks was 87.5%. The forecasted values were within the 95% confidence band of the actual ending 52 week data, though at the high end of the band. ConClusions: The SARIMA model was an appropriate predictor for flu cases in 2013-4. The data used to construct the model included flu epidemics so removing these time periods would result in a model more appropri-ate for non-epidemic periods.

PRM103develOPMent OF A glObAl eCOnOMiC MOdel tO evAlUAte tHe COst-eFFeCtiveness OF tARgeted tReAtMents Using COMPAniOn diAgnOstiCs in AdvAnCed/MetAstAtiC CAnCeR tReAtMentMathurin K., Beauchemin C., Lachaine J.

University of Montreal, Montreal, QC, CanadaobjeCtives: With the development of high priced new targeted treatment for can-cer, there is a need to know as soon as possible if these treatments are likely to be cost-effective. The objective of this study was to develop a model with global parameters to estimate the cost-effectiveness of targeted treatments using compan-ion diagnostics in advanced/metastatic cancer treatment. Methods: The model was developed to take into account parameters usually considered in conventional economic models in cancer (treatment costs, costs of cancer care, target population characteristics, survival data, utilities, disutilities and costs associated with adverse events (AEs), etc.), and also parameters specific to the companion diagnostic itself (mutation prevalence, test specificity and sensitivity, and cost). The model had to allow performing cost-utility analyses from both a Health Ministry and a societal perspective and for most common cancers (lung, breast, colorectal, prostate, cer-vical/endometrial, bladder, and non-Hodgkin’s lymphoma). Results: The global model comprises a decision tree and a lifetime Markov model. The decision tree takes into account the sensitivity and specificity and cost of the companion diag-nosis, and the prevalence of the biomarker/mutation in the eligible population. The Markov model with monthly cycles includes the following 3 health states: progression-free, progressive disease and death. Intrisic parameters of the model comprise the mean characteristics of the target population, utilities associated with health states, disutilities and costs associated with AEs, and costs associated with drug administration, cancer care, end-of-life care, follow-up visits, productivity

A562 VA L U E I N H E A LT H 1 7 ( 2 0 1 4 ) A 3 2 3 – A 6 8 6

PRM109visUAlizing MetHOds FOR disCRete-event-siMUlAtiOns Using tHe exAMPle OF A bReAst CAnCeR deCisiOn-AnAlytiC MOdelJahn B.

1, Rochau U.

2, Shterjovska J.

1, Kurzthaler C.

3, Kluibenschädl M.

1, Urach C.

4, Einzinger P.

4, Piringer H.

5, Popper N.

4, Siebert U.

6

1UMIT - University for Health Sciences, Medical Informatics and Technology, Hall in Tyrol, Austria, 2UMIT - University for Health Sciences, Medical Informatics and Technology/ ONCOTYROL - Center for Personalized Cancer Medicine, Hall in Tyrol/ Innsbruck, Austria, 3UMIT - University for Health Sciences, Medical Informatics and Technology / Oncotyrol - Center for Personalized Cancer Medicine, Hall i. T. / Innsbruck, Austria, 4Vienna University of Technology, dwh Simulation Services, Wien, Austria, 5VRVis Zentrum für Virtual Reality und Visualisierung Forschungs-GmbH, Vienna, Austria, 6Medical Informatics and Technology, and Director of the Division for Health Technology Assessment and Bioinformatics, ONCOTYROL, Hall i. T, AustriaobjeCtives: Discrete-Event-Simulation (DES) is a commonly used modeling tool to analyze the comparative effectiveness of alternative health technologies and to optimize resource allocation in health care settings. DES models are often rather complex and visualization is very important to improve transparency and acceptability. This study aims to illustrate and contrast alternative visualization techniques on a decision-analytic model for breast cancer. Methods: DES visu-alization methods and their applications in health care, engineering, and opera-tions research were sought from a wide variety of sources, including literature databases (e.g., PubMed) and webpages of simulation conference (e.g., Winter Simulation Conference), academic societies etc. Based on this review, alternative visualization techniques for the conceptual model were selected, applied on a real world modeling example and compared. Results: In health care, the recently published ISPOR-SMDM Modeling Good Research Practice guidelines recommend flow diagrams or state charts to represent the key elements of a model, including the possible pathways, and the presence of queues and decision points. For flow charts, we found an international standard (ISO 5807). The application of standards like this could support harmonization of process-oriented models. In general, flow charts may lack the information of health states and transitions between health states that are relevant for clinicians to review the model. The semantic for state charts invented by Harel provides a further development of the bubble diagrams of State-Transition (Markov) Models (e.g. one state containing other states, one state detects changes in another). In state charts, health states could explicitly be named but treatment processes and resources use are less explicit. For DES software implementation, state charts seem to be less intuitive. For both methods, the application of visualization standards and guidelines was not always straight forward for our breast cancer model. ConClusions: In the case example there was no superior visualization technique.

PRM110MiCROsiMUlAtiOn MOdel FOR tHe AssessMent OF PeRsOnAlized CAnCeR CARe: tHe MAPCCA MOdel FRAMeWORkVan der Meijde E.

1, van den Eertwegh A.J.

1, Fijneman R.J.

1, Meijer G.A.

1, Linn S.C.

2, Coupe V.M.

1

1VU University Medical Center, Amsterdam, The Netherlands, 2Netherlands Cancer Institute, Amsterdam, The NetherlandsobjeCtives: Most cancer care models are based on observed clinical events such as recurrence-free and overall survival. Times at which events are recorded depend not only on effectiveness of treatment, but also on timing of examinations and types of tests performed. Should these change, observation times would change as well. Construct a microsimulation model that describes the cancer disease process using a description of underlying tumor growth as well as its interaction with diagnostics, treatments and surveillance. The aim is to arrive at a frame-work that allows for exploration of the impact of simultaneously altering two or more aspects of the care process. Methods: The framework consists of two components; the disease model and the clinical management module. The disease model consists of atumor level, describing the growth and metastasis of the tumor, and a patient level, describing clinical observed states, such as recurrence and death, either from the disease or other causes. The clinical management module consists of the care patients receive, i.e. the diagnostic process, treatment and surveillance. This module interacts with the disease process, influencing the rate of transitioning between tumor growth states at the tumor level, and the rate of detecting a recurrence at the patient level. Results: A simulation study was performed to examine the feasibility of applying the framework to melanoma progression. Results demonstrated stage specific recurrence rates similar to those found in literature. ConClusions: The proposed microsimulation model frame-work allows for generating individual patient histories by simulating underlying tumor growth in interaction with clinical management. Our modeling approach allows for the exploration of the potential of drugs intervening in different parts of the tumor growth pathway. In addition, the approach allows for the evaluation of changing diagnostic patterns.

PRM111MetHOdOlOgiCAl evAlUAtiOn OF tHe iMPACt OF sURvivAl COsts in OnCOlOgy MOdellingTaylor M.

1, Filby A.

1, Proudfoot C.

2

1York Health Economics Consortium, York, UK, 2Sanofi, Guildford, UKobjeCtives: Economic evaluations typically include all costs relevant to a disease, not only drug-related costs. This is particularly relevant to oncology modelling, as costs are assigned to each health state in the model, and, therefore, extending sur-vival also increases costs. Because patients often incur higher health care costs in the post-progressed state of disease where costs of disease management are high, extending survival and increasing a patient’s time in the post-progressed stage can be particularly costly. Empirical analyses of the implications of such methods have not yet been extensively investigated by assessing different scenarios such as baseline severity and prognosis. The objective of this research was to investigate the methodology used in oncology modelling, and to determine the effect that this has on predicted cost-effectiveness. Methods: We developed a flexible three-state

Zidovudine + Lamivudine combination therapy vs Zidovudine monotherapy, to treat HIV infection. Based on probabilistic simulations, cumulative incremental net monetary benefits (CINMB) at a CE threshold of £20,000/QALY and probabili-ties of being cost-effective at various time-horizons (1-20 years) were estimated. Further, for each time-horizon, a CINMB frequency distribution was plotted and summary statistics were estimated. Results: For the combination therapy, while the outcome uncertainty increased over time, the decision uncertainty decreased. 95% confidence interval for expected CINMB was narrowest at year 1 (-1,771£ to -1,755£) and widest at year 7 (2,101£ to 2,209£); simultaneously the probability of being cost effective increased from 5% to 80% during this time. Outcome uncer-tainty, measured as the standard deviation of CINMB values stabilized after 5 years while probability of the combination therapy being cost effective continued to increase, indicating that decision uncertainty does not vary in tandem with outcome uncertainty. ConClusions: The above analysis shows that higher out-come uncertainty does not necessarily lead to higher decision uncertainty. CINMB could be a useful tool to observe the relationships between outcome uncertainty, decision uncertainty and time.

PRM107develOPMent OF A MOdel tO Assess tHe COst-eFFeCtiveness OF tHeRAPies FOR PAtients WitH tyPe 2 diAbetes MellitUs (t2dM) FOllOWing A ReFeRenCe MOdel FRAMeWORkAguiar-Ibáñez R.

1, Palencia R.

2, Kandaswamy P.

3, Flavin J.

4, Gauthier A.

5, Davies M.J.

6

1Amaris Consulting UK, London, UK, 2Boehringer Ingelheim GmbH, Ingelheim am Rhein, Germany, 3Boehringer Ingelheim UK, Bracknell, UK, 4Boehringer Ingelheim Canada Ltd, Burlingon, ON, Canada, 5Amaris, London, UK, 6University of Leicester, Leicester, UKobjeCtives: To describe the practical approach implemented to construct a global cost-effectiveness model for T2DM therapies following a framework pro-posed for the development of reference models to inform public funding deci-sions. Methods: 1) A systematic review of published models was conducted to conceptualise the model in terms of natural history and relevant effects to include. 2) Clinical and health economic experts were selected to provide feedback during the model conceptualisation (to identify the appropriate modelling technique), the model implementation and the assessment of the results. 3) The model was built and populated based on the systematic identification of best available data, a net-work meta-analyses, a review of previous T2DM submissions to health authorities and other published information. The model incorporated several structures for uncertain areas, such as: treatment patterns; type and timing of adverse events; their impact in the occurrence of long-term complications; and the impact of weight changes on relevant endpoints. 4) The model was then validated based on out-puts’ accuracy, feedback from country affiliates and consistency with the CORE model results. 5) The critical feedback received by HTA bodies has also been used to refine the model and improve its credibility accordingly. Results: Experts’ input proved invaluable at each developmental stage. One challenge related to the com-parability with other published T2DM models, which were not fully transparent regarding assumptions. This framework resulted in a flexible model, accurate and stable, and easily adaptable to different health care systems. Country adaptations have contributed to the identification of aspects that require relevant structural changes and their rationale. ConClusions: The followed framework enhanced the transparency of the model and the accuracy of the results. Using a reference model across different countries, with adaptations made in consistency with this model, should help ensure consistent and comparable evaluations of the model across different countries.

PRM108Assessing tHe RelAtiOnsHiP betWeen individUAl AttRibUtes identiFied in RevieW OF MUlti-CRiteRiA deCisiOn AnAlysis (MCdA) OF RARe diseAses And AnnUAl tReAtMent COsts in RARe endOCRine disORdeRsSchey C.

1, Irwin J.

2, Teneishvili M.

2, Krabbe P.F.M.

3, Connolly M.

4

1University of Groningen, St Prex, Switzerland, 2Shire Pharmaceuticals, Maidenhead, UK, 3University of Groningen, University Medical Center Groningen, Groningen, The Netherlands, 4University of Groningen, Groningen, The NetherlandsobjeCtives: Payers have a perception that orphan products are extremely expensive. The current health technology assessment (HTA) systems might be too restrictive for orphan drugs, therefore potentially denying patients access to life-saving medicines. While price is important, it should be considered in relation to a broader range of product attributes, such as unmet need and disease severity that are not considered in cost-effectiveness analysis used by many HTA agencies. To overcome these challenges multi-criteria decision analysis (MCDA) has been proposed as an alternative to evaluate technologies. The aim of this study was to identify criteria reported in the literature, and to assess their impact on the total “score” for each product in relation to price. Methods: A systematic literature review was conducted to identify the most frequently cited attributes in MCDA. From the leading attributes identified, we reviewed and plotted the relationship between single attributes and the average annual treatment costs for several drugs used in the treatment of endocrine-related rare diseases. Annual treatment cost was based on UK prices for the average daily dose per patient. Results: The three most frequently mentioned attributes were ‘disease severity’, ‘treatment impact on condition’, and ‘level of research undertaken to support use of the prod-uct’. Disease severity was not shown to influence product price. Similarly, orphan drugs are not necessarily more expensive than products without orphan drug status. There is little discernible relationship between treatment ‘convenience’ and average annual treatment cost. A trend was observed between the market size and the average annual treatment cost. ConClusions: If society is concerned about equity and equal access to medicines for all patients, MCDA may offer a viable alternative to inform in reimbursement decisions for orphan drugs. The analysis can be used to inform investigations on the application of MCDAs in rare diseases.