Pavel Fiser (1), Daniel Kaschek (2), Henning Schmidt (2), Anna Georgieva Kondic (1)
(1) Economic and Data Sciences, MSD, Rahway, NJ, USA, (2) IntiQuan, Basel, Switzerland
Objectives: There were two main, related, objectives of our study. The technical goal was to create a comprehensive individualized and transparent modelling framework that could be easily used to evaluate the potential for economic differentiation of new drug candidates against existing therapies. An example application of particular interest was the evaluation of a novel anticoagulant being considered for stroke prevention. This example comes from a therapeutic area that requires large and costly outcome trials and can benefit from informed early go/no go decisions. Additionally, this modelling study evaluated the handling, user-friendliness, and efficiency of model development and readability of the resulting model as compared to existing tools.
Methods: A Markov cohort [1] economic model described in Verhoef at al [2] compares three different factor X molecules and coumarin derivatives in the settings of the UK and the Netherlands. The model considers a uniform group of patients with atrial fibrillation who can experience a variety of cardio-vascular (CV) events, such as myocardial infarction, systemic embolism, ischaemic stroke, transient ischaemic attack, as well as bleeding events, each of which carries a certain risk for disability or death. The comparison between the different therapies is based on their published efficacy, safety, effect on quality of life and costs. An initial model using the mean parameters as presented in [2] was first implemented in heRo3 [3], a web-based tool that runs the R-based economic modelling package heemod [4]. The Markov formulation of the model in heRo3 required a large number of health states which, combined with the different therapies made it cumbersome to characterize the different combinations of incidents, especially if we were to consider different patient sub-populations. The model becomes difficult to understand by the different stakeholders (reviewers, non-modeller partners). A more intuitive representation of such a model is enabled by the Discretely Integrated Conditions and Events (DICE) methodology [5], where states are not mutually exclusive but each patient simultaneously carries several attributes characterizing the patient’s health state. This DICE methodology forms the basis of the R based Health Economics Modelling (RHEM) Framework, developed for this work. The example model was implemented, simulated and analyzed in the RHEM Language using the DICE methodology.
Results: We developed the RHEM Framework and subsequently formulated the example model in the new RHEM Language. This allowed us to represent all required features for an example compound and implement the desired target product profile while also considering the uncertainty associated with early clinical development and account for the variability in the patient population. The resulting RHEM model code is well readable and intuitively understandable, supporting efficient model building and understanding by reviewers. The simulations in the RHEM Framework reveal the separation between different groups of patients based on their covariate characteristics and suggest that enriching for the responder patients can increase the probability of success for a clinical development program. In addition, the deterministic and probabilistic sensitivity analyses point out to the most sensitive parameters in overall model predictions. As such, the new RHEM modelling tool can be used to inform future clinical trials to gather the data necessary to inform the sensitive parameters.
Conclusions: The RHEM modelling framework, as applied to the novel anticoagulant agent demonstrates that it is possible to allow for efficient and user-friendly definition of Health Economic models that integrate diverse information (pharmaceutical properties, efficacy, safety, patient characteristics and health resource utilization) in order to characterize the value proposition of new therapeutic molecules and target populations who may benefit from new treatments. The resulting RHEM modelling approach is intuitive to understand, can scale up to large problems and can be applied to different therapeutic areas in order to inform clinical development and registration strategy for molecules under development. As the underlying RHEM simulator is based on the DICE methodology, different modelling types, including Markov models and Partitioned Survival models [6] can be supported and the modeler can choose the methodology that is most useful for a particular problem and therapeutic area.
References:
[1] Beck JR, Pauker SG. The Markov process in medical prognosis. Medical decision making. 1983 Dec;3(4):419-58.
[2] Verhoef TI, Redekop WK, Hasrat F, de Boer A, van der Zee AHM. Cost effectiveness of new oral anticoagulants for stroke prevention in patients with atrial fibrillation in two different European healthcare settings. Am. J. Cardiovasc Drugs. 2014; 14:451-462.
[3] heRo3, http://hero3models.info/
[4] (https://cran.rproject.org/web/packages/heemod/indelx.html)
[5] Caro JJ, Möller J, Getsios D. Discrete event simulation: the preferred technique for health economic evaluations?. Value in health. 2010 Dec 1;13(8):1056-60.
[6] Williams C, Lewsey JD, Mackay DF, Briggs AH. Estimation of survival probabilities for use in cost-effectiveness analyses: a comparison of a multi-state modelling survival analysis approach with partitioned survival and Markov decision-analytic modelling, Medical Decision Making, May, 2017.
Reference: PAGE 27 (2018) Abstr 8661 [www.page-meeting.org/?abstract=8661]
Poster: Drug/Disease Modelling - Other Topics