MVEX01-17-20 Monte-Carlo simulation in pharmaceutical decision making

Captario AB was started in 2012 with a mission to improve decision making in pharmaceutical companies. By understanding trade-­‐offs between cost, time, risk and value, companies can make decisions in the right context. This leads to less sub-­‐optimal decisions and a better understanding of the effect of decisions. Captario develops SUM – Strategic Uncertainty Management. SUM is a software agent allowing users to model drug project activities that eventually lead to a product launch. In the tool, the user enters estimates for time and cost for each activity, and also provides quantified estimates of project risk (figure 1). SUM uses Monte-Carlo simulation to sample the vast outcome space that such a model defines.

A drug project can be divided into three main parts: the drug discovery, the drug development, and the market phase. The drug discovery is the research phase, where the company discovers new drug entities. In the drug development the molecule undergoes testing in animal models and clinical trials. The market phase the product enters a market place and generates a return. The second and third phases are focus areas for Captario SUM. Using drug development as an example here, it is a stage-gate process that generally consist of preclinical testing, followed by three clinical phases and then a regulatory approval process. It is a highly regulated process. After each stage in the process, project stakeholders will look at the evidence generated, and determine the fate of the project. Either to continue or to stop the development.

 
Captario SUM will use the model in a simulation, and will generate information about the possible outcomes of the project and will also generate key financial metrics, such as ROI, NPV expected gain, expected loss etc. A real project example is generally a bit more complex than this example, but shares some attributes

  • There are dependent, and independent variables.
  • The dependent output variables (the key financial instruments) are important to compare projects.
  • There can be chains of dependencies between particular variables and the key financial instruments.
  • The simulation output data is available for analysis.

  •  
    The problem 
    We would like to further develop the analyses on how variables, i.e., user assumptions, independent and otherwise, affects project cost, risk and value, and order all input variables based on how much they effect output. We would like to investigate how multivariate methods and perhaps machine learning methods that can be used and tailored to determine the correlation between inbound variables and the major dependent variables in our stochastic drug project models. With basic sensitivity analysis, a huge number of simulation iterations are needed to generate a converging result. The number of iterations has to increase with increasing number of variables in the model. We want to investigate other methods of analyzing the effect, without necessarily increasing the number of simulation iterations. 

     
    Obs! För GU-studenter räknas projektet som ett projekt i Matematisk Statistik (MSG900/MSG910).
     
    Projektkod MVEX01-17-20
    Gruppstorlek 3-4 studenter
    Förkunskapskrav: grundläggande sannolikhetsteori, stokastiska processer
    Handledare Magnus Ytterstad, magnus.ytterstad@captario.com​
    Examinator Maria Roginskaya, Marina Axelson-Fisk
    Institution Matematiska Vetenskaper

     

    Publicerad: må 31 okt 2016. Ändrad: on 09 nov 2016