AN INTRODUCTION TO PHARSIGHT DRUG MODEL EXPLORER (DMX ) WEB SERVER Software to Visualize and Communicate Model- Based Product Profiles in Clinical Development White Paper July 2007 Pharsight Corporation 321 East Evelyn Avenue, 3 rd Floor Mountain View, CA 94041 Tel: 650-314-3800 Copyright 2007 Pharsight Corporation All Rights Reserved - 1
Introduction Drug development professionals need new visualization and communication tools to understand a compound s behavior and expected performance in order to make important decisions. Pharsight Drug Model Explorer (DMX ) software fills this need by allowing drug development project team members to easily view and query information about likely compound product profiles generated from quantitative drug models. In a recently issued guidance, the FDA has demonstrated its interest in target product profiles as strategic communication tools 1. FDA officials have also continued to publicly discuss the potential for model-based drug development methods to capture and communicate important information about a clinical development program, and as a means of strategically developing target product profiles 2. Through the DMX interface any team member, from technical experts to clinical project leaders, can compare probabilistic outcomes for different endpoints, treatment strategies, patient populations, and competing products. Results are presented as a series of plots, tables and summary indicators of drug success against performance targets that collectively define a product profile. These plots, tables and profile performance summaries can be quickly updated based on new underlying data and simulated outcomes based on models of clinical effect. DMX results are accessible from networked desktop or laptop computers for individual exploration, interactive team discussion, and communication of development strategies and program alternatives with senior decision-makers. DMX facilitates answers to strategic development questions, such as: What is the expected clinical response for a treatment strategy in a particular patient population? What is the level of certainty surrounding predicted response? How do different treatment strategies and target patient sub-populations impact response? What is the probability that response is less or greater than a specific target? What dose is required to achieve a target response? What is the probability of achieving a specific efficacy target while keeping probability for adverse events below a certain level? How do the attributes for the compound compare to competitors? What is optimal positioning strategy versus competitors to balance safety and efficacy? This white paper introduces DMX Web Server and describes how it can be used to improve decision-making by performing dynamic what-if analysis on a compound s likely clinical potential. The white paper also describes how DMX lowers technical barriers to understanding and use of model-based results. Finally, this paper describes how DMX provides a framework for accessing and communicating complex model-based information, in the context of clinical product profiles, that can be leveraged across drug programs to make consistent comparisons of development program alternatives. Copyright 2007 Pharsight Corporation All Rights Reserved - 2
The Information and Communication Challenge Drug development is a dynamic information gathering activity. New information emerges not only from studies conducted on the drug under development but also from studies on competing products for a similar or related indication. Drug and disease models quantitatively integrate all the available pre-clinical and clinical data on a compound, public data in the clinical and scientific literature, and expert opinion to inform an understanding of compound attributes and behavior. These probabilistic mathematical models can then be used to run simulations that predict, within an interval of uncertainty, the expected clinical results for a wide range of treatment and patient scenarios involving the compound and its competitors. This means that far earlier in the process than usual, a development team has significant quantitative and objective insight into the likely behavior of a compound compared to its competitors. This, in turn, provides the team a powerful basis for making good decisions about their development program. The FDA has recognized systematic application of model-based drug development as an important approach to improving drug development knowledge management and development decision-making 3. For project team members evaluating the likely product profiles of a compound in clinical development, however, accessing and using complex information contained in drug-disease models poses significant challenges. First, team members approach key development questions from their own functional and disciplinary vantage points (e.g., pharmacokinetics/pharmacodynamics (PK/PD), clinical medicine, biostatistics, clinical pharmacology, marketing, regulatory) and often have limited time and means by which to apply their individual and collective experience against, and develop confidence in, modeling and simulation results. Second, modeling and simulation results are themselves large matrices of numbers that require preparation by modeling experts into static plots and tables for team consumption. If the plots and tables don t fully address team questions, new questions arise, or as new data becomes available the modeling expert must run new simulations and return with new plots and tables. This process is time-consuming and limits the team s ability to engage in what-if analysis and dynamic discussion of alternative development scenarios. Finally, different team members on geographically distributed project teams may be exposed to different sources of information that are not integrated and generally cannot be directly compared. Data on the compound under development and its competitors is typically scattered across public and proprietary databases. As new information becomes available and is integrated into existing drug models, teams need a common collaboration mechanism for developing a holistic picture of compound effects. The absence of a visual language for viewing modeling and simulation results limits a team s ability to reach consensus, in a quantifiable way, on what is certain and uncertain about the compound. Copyright 2007 Pharsight Corporation All Rights Reserved - 3
Drug Model Explorer Web Server Overview Drug Model Explorer (DMX) Web Server is a software application for improved team communication and assessment of likely product profiles. At its core, DMX is a webbased visualization and communication tool to assess model-based compound attributes and uncertainty versus competing therapies. Through the DMX web interface, clinical development project team members are able to compare probabilistic outcomes for different endpoints, treatment strategies, patient populations, and competing products. For a given set of input assumptions, outputs are displayed in the form of graphical dose-response plots and tables, and as compact summaries of predicted success against targets that define a product profile (see Figure 1). DMX output can be copied or downloaded into compatible Windows-based applications, such as Microsoft Word and PowerPoint. As users query model-based results to dynamically address key development questions, DMX enables enhanced understanding of a compound s likely clinical potential. Figure 1: DMX Product Profile Summary. DMX is a web-based visualization tool that helps drug development project teams easily view and query model-based product attributes generated from quantitative drug models. For a given set of input assumptions, results are displayed as plots, tables and product profile performance summaries. The picture above shows the probability of a 100mg dose of the hypothetical schizophrenia compound Alanapine reaching target values for selected safety/efficacy endpoints in four product profiles: three team profiles ( Do Not File, Acceptable, Ideal ) and one personal profile ( Competitive ). Copyright 2007 Pharsight Corporation All Rights Reserved - 4
Drug Model Explorer offers solutions to the information and communication challenges facing project teams and decision-makers: Dynamic What-If Analysis on Compound Behavior for Decision-Making DMX is a flexible tool for helping project teams build consensus on likely product profiles, key tradeoffs, and new information on competing products. DMX provides ready access to a database of pre-simulated responses within a given decision domain of inputs and scenarios (e.g., endpoints, doses, covariates, comparator treatments) defined by the team as part of the modeling process. Doses of interest for which predictions have not already been simulated are interpolated. The ability to answer what-if questions in a team meeting, for example, is often difficult to achieve with a standard slide presentation. Raw model-based simulation output is packaged and stored for use by DMX in a file format designed to accommodate large datasets that satisfy the model exploration needs of the team across many input scenarios at any stage of development. Lower Technical Barriers to Accessing Model-Based Results DMX removes the need for project team members to familiarize themselves with detailed modeling techniques in order to grasp the key insights provided by modeling and simulation. DMX allows direct interaction with modeling results. Queries are selected and viewed by individual team members, as opposed to expert modelers, to help them develop intuition of and appreciation for modelbased results. Greater comfort with the modeling results increases confidence that a drug model accurately represents important drug characteristics and increases trust in the insights provided by quantitative models. Knowledge Management Within and Across Development Programs DMX Web Server supports secure management of model-based data for on-thespot access and communication, and can be easily updated when new information becomes available or to answer new questions. The DMX user interface represents complex models and large amounts of simulated data, facilitating development of a holistic picture of compound effects. The DMX visual presentation structure is applicable across drug programs, allowing teams to share information rapidly and clearly and to make consistent comparisons of development program alternatives. Copyright 2007 Pharsight Corporation All Rights Reserved - 5
DMX Web Server Components and Functionality The DMX Web Server system allows end-users to view simulated drug model outputs from browser-based client software that communicates with a secure server over customer local- and wide-area networks (see Figure 2). The DMX web client can also be used over the Internet using standard protocols or Virtual Private Network (VPN) technologies. Drug-disease model outputs and the metadata describing these outputs reside on a secure central server. Browser-based administrative tools are used to manage DMX users, projects, and project-based system access permissions. Desktop publishing tools convert pre-simulated drug model outputs into a binary format for uploading to (or from) the server. The first step of the publishing process requires drug model data outputs to be saved in S-PLUS or SAS file format, although DMX offers flexibility for modelers to run the underlying simulations using these or other industry standard software packages such as NONMEM. A DMX desktop client allows users to view drug model outputs while detached from the network on their local or laptop machines (e.g., to support discussions with regulatory authorities). Once installed, the DMX desktop client application also requires a copy of DMX data and metadata, which in this case do not reside on a central server but are available for local access. The DMX publishing tools support data preparation for access by either the web-client or desktop application. Figure 2: The DMX Web Server System. The DMX system consists of browser-based client software for viewing simulated drug-model outputs that reside on a secure central server. Desktop publishing tools prepare and upload data for use by the web client or for local access via a desktop client. System administration tools are accessed via the web client. Copyright 2007 Pharsight Corporation All Rights Reserved - 6
Together, these components support important DMX system functionality: User-Friendly Input Controls and Output Display Point-and-click controls on the DMX web client interface allow users to select different inputs for analysis and to specify an uncertainty (confidence) interval for a given set of inputs. Results are displayed in both graphical and tabular format. Available inputs based on the underlying models include continuous/categorical endpoints and covariates, drug-dose selections for the compound under development, and reference (competitor) compounds (see Figure 3). New profiles functionality extends previously available ranges functionality, and allows users to more directly test the likelihood that a compound will meet its target product profile based on specific performance criteria set for one or more endpoints (Figure 1). Figure 3: DMX Dose-Response Interface. Users can calculate the probability that a given treatment will achieve a target endpoint range(s). The picture above compares response for Alanapine 100mg, 150mg and 180mg on the endpoint PANSS-Neg in the context of three defined performance ranges: No File, Acceptable, and Target. Alanapine 100mg has low probability of Target efficacy versus competing (reference) treatment Lobotacain 400mg (i.e., only 20% of the response distribution is in the defined Target range), but the probability of achieving Target performance at higher Alanapine doses appears more likely. Copyright 2007 Pharsight Corporation All Rights Reserved - 7
Links to Model Assumptions and Documentation The DMX interface provides quick reference to supporting drug model documentation. By clicking on the View Model Documentation button (Figure 3), users can view documents created in any Windows -based application file format. This helps provide teams with increased confidence in DMX results by being able to review model equations and assumptions, validate source data, and discuss implications for decision-making without ever leaving the DMX software. Desktop Publishing Tools DMX publishing tools provide the essential link between model-based simulation output data and the visual presentation of that data in the DMX web interface. These easy-to-use desktop tools convert and package raw simulation outputs into a data format that is loaded, interpreted and displayed by the DMX web and desktop clients. The publishing tools include utilities for the modeler to edit model metadata (endpoint names, data labels) for clear communication with those who were not involved in the technical aspects of running the simulations. Resulting data and metadata files are then published to the central web server for access by the team via the web client (see Figure 4). Figure 4: DMX Publishing Tools. Desktop tools convert and prepare pre-simulated drug model data and associated files for use by the DMX web client. Copyright 2007 Pharsight Corporation All Rights Reserved - 8
Project and Version-Based System Management Browser-based administrative tools manage projects, permissions and users within the DMX system. DMX data is organized as a series of projects and corresponding project versions. A project can hold multiple project versions, which are created with the publishing tools and reside on the central server for access by the team. Only users with explicitly assigned permissions can view projects or specific versions of projects, publish versions to the server, or perform other system management tasks. Creation and Management of Response Views Pre-defined DMX dose-response views can be created using the desktop client, and included by the modeler when publishing to the web server. This functionality allows project team members to immediately access plots and (optional) tables that highlight important aspects or issues of the development project, and also provides a historical record of team discussion and progress. Publishers can choose to denote any one of the views they prepare as the default view. Plot overlay controls on the DMX interface support the ability for both pre-defined views and ad hoc queries to include multiple variables of interest, such as simultaneous dose-response for a safety and efficacy endpoint and their associated covariates (see Figure 5). Figure 5: DMX Response Views. Each project version can include pre-defined views of interest to the team, including a default view. The modeler creates these views during the publishing process. The picture above shows the overlay view PANSS and Weight, which allows users to simultaneously assess the dose-response for Alanapine on the efficacy endpoint PANSS and the safety measurement Wt. Gain, stratified by different treatment covariates (e.g., gender). Copyright 2007 Pharsight Corporation All Rights Reserved - 9
Creation and Management of Profiles Profiles represent target values for one or more modeled DMX endpoints that can be created using either the desktop client or web client. This functionality allows project team members to define endpoint performance targets on an absolute or relative basis (e.g., versus a competitive reference treatment), and to assemble them in the context of a product profile. The profile view presents a compact summary of the relationship between desired targets and modeled endpoints, and displays the chance that the drug effect will meet the defined target for a given set of input or treatment scenario assumptions (see Figure 6). There are two types of profiles: team and personal. Team profiles are created by DMX publishers and are available on the web client to summarize competitive drug performance and provide a ready basis for further team exploration. Personal profiles are created by individual DMX web client users, either as copies of team profiles or as newly defined profiles. Using the web client, team and personal profiles can also be imported from other published versions of the same project (Figure 1). Figure 6: DMX Profiles. DMX profiles represent a collection of success criteria defined for one or all modeled endpoints, and display the probability that a drug will meet desired targets. Team profiles are created by modelers and published to the web server for team review. Personal profiles are created and managed by DMX web client users. The picture above shows the personal profile called Competitive, for which users can modify endpoint performance targets, test different input scenario assumptions (e.g., dose, covariates) for Alanapine or other treatments, or explore more detailed dose-response data. Copyright 2007 Pharsight Corporation All Rights Reserved - 10
Publishing and Exploring Product Profiles With DMX The annotated workflow illustration in Figure 7 provides a specific context for how DMX is used on a development project and for the potential interplay between project team members, technical modeling experts, and decision-makers. 1. Drug Model Building and Simulation The project team builds drug-disease models and performs associated simulations on the basis of available information and an understanding of key development questions. 2. Publish Drug Model Results to DMX DMX is populated by the project team with a database that contains simulated probability distributions of efficacy, safety, or other endpoints measures as a function of specific model inputs such as treatment options (drug, dose, dose frequency, etc.), patient populations, and assumptions. The database and its associated definitions, which are defined by the modeling expert with the team as part of the modeling process, comprise the range of scenarios that can be explored by DMX. Project modelers also populate DMX with an overview of the model pedigree (documentation on source data, validation, conclusions). DMX publishing tools are used to upload the data, metadata, and model documentation, along with any specified dose-response views, team profiles and profile settings, to the server for team exploration. Figure 7: Collaborative Team Workflows Using DMX. The illustrated workflow above provides a context for the interplay between drug model building and simulation, visualization of simulated product attributes and uncertainty using DMX, and DMX-supported communication processes with project teams, modeling experts, and senior decision-makers. Copyright 2007 Pharsight Corporation All Rights Reserved - 11
3. Team DMX Exploration, Model Refinement and Feedback Project team members use DMX to actively explore simulated model outputs of drug product attributes and associated uncertainty. The range of exploration and comparisons includes summary predictions of drug success against performance targets that collectively define a product profile, and underlying dose-response for different endpoints, different treatment regimens, patient populations, or different competing products. Because model building and decision-making are interactive processes, new questions will arise, assumptions will change, new data will become available, or certain questions will become obsolete. As a result, models are updated and/or simulated results are efficiently re-published to the server for ongoing team review and consideration. 4. Frame Insights and Recommendations Project teams use certain DMX views and profiles that capture the main insights from their exploration of the model to support key development program recommendations. 5. Communication and Decision-Making DMX is used to communicate the effects of the compound in development relative to internal and external competitors. Flexible exploration enhances the project team's ability to make informed decisions regarding product positioning and development strategy. Senior decision-makers are presented with views and profiles that summarize the rationale underlying team recommendations, and will have the option to modify certain choices themselves, using the summary views and profiles as a starting point. Copyright 2007 Pharsight Corporation All Rights Reserved - 12
DMX Web Server Technology DMX Web Server has a three-tier architecture to support stability and scalability requirements. This architecture is implemented using software and database components compatible with pharmaceutical industry standards (see Figure 8). Figure 8: DMX Web Server Architecture. The DMX Web Server architecture is implemented with components compatible with pharmaceutical industry standards. DMX uses Active Directory for authentication of named users to the server. Access to the web client is controlled through Active Directory plus application-level permissions established by the project team for viewing and publishing. The DMX web client runs in Internet Explorer, allowing for ease of deployment. DMX Web Server relies on the following software components: Front-End Layer MS Windows 2000 (SP1, 2, 3, 4) or Windows XP (SP1, 2) Server/Middle Layer Windows Server 2003 Internet Information Services 6.0 MS.NET Framework 1.1 Internet Explorer 5.5, 6.0, 7.0 Adobe Acrobat Reader MDAC 2.8 Active Directory Data Layer Oracle 9i or 10g Copyright 2007 Pharsight Corporation All Rights Reserved - 13
Benefits of Drug Model Explorer DMX supports Pharsight s mission to improve clinical drug development decisionmaking using an integrated, quantitative, model-based methodology. With DMX, companies can continue to capitalize on model-based drug development to accelerate clinical development timelines and cut development costs by making program decisions earlier and with greater confidence. The key benefits of deploying DMX within a drug development enterprise are: Ability to rapidly answer what-if questions on product attributes Lower technical barriers to understanding and using model-based results Knowledge management within and across development programs Advanced product technology, scalability and ease of deployment Contact Information For further information regarding Drug Model Explorer please contact: James Hayden Nancy Risch Senior Vice President, Global Sales Vice President, European Sales jhayden@pharsight.com nrisch@pharsight.com 650.314.3760 781.771.2243 About Pharsight Corporation Pharsight Corporation (OTC Bulletin Board: PHST) develops and markets integrated products and services that enable pharmaceutical and biotechnology companies to achieve significant and enduring improvements in the development and use of therapeutic products. The company's goal is to help customers reduce the time, cost and risk of drug development, as well as optimize the post-approval marketing and use of pharmaceutical products. Pharsight's approach enhances the fundamental element of drug development success: strong decision-making. By adopting the Pharsight approach, customers acquire a new decision-making process with the potential to systematically improve every level and phase of their business and scientific processes. Pharsight Corporation is headquartered in Mountain View, California. Information about Pharsight is available at http://www.pharsight.com. All contents Copyright 2007 Pharsight Corporation. All rights reserved. The copyright for this document is owned by Pharsight Corporation. Pharsight, Drug Model Explorer and DMX are registered trademarks of Pharsight Corporation. All other brand and product names are trademarks or registered trademarks of their respective holders. Copyright 2007 Pharsight Corporation All Rights Reserved - 14
References 1 FDA Guidance for Industry and Review Staff: Target Product Profile A Strategic Development Process Tool. Food and Drug Administration, March 2007. Available at: http://www.fda.gov/cder/guidance/6910dft.htm 2 Lesko LJ. Paving the Critical Path: How Can Clinical Pharmacology Help Achieve the Vision? Clinical Pharmacology and Therapeutics, 81:2. February 2007. 3 Innovation or Stagnation: Challenge and Opportunity on the Critical Path to New Medical Products, Food and Drug Administration, March 2004. Available at: http://www.fda.gov/oc/initiatives/criticalpath/ Copyright 2007 Pharsight Corporation All Rights Reserved - 15