A Standard Computable Clinical Trial Protocol: The Role of the BRIDG Model

Transcription

1 CLINICAL RESEARCH STANDARDS 383 Cara Willoughby, MS Eli Lilly and Company, Indianapolis, Indiana Doug Fridsma, MD, PhD University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania Lisa Chatterjee, MS Digital Infuzion, Inc., Gaithersburg, Maryland John Speakman National Cancer Institute, Rockville, Maryland Julie Evans Clinical Data Interchange Standards Consortium, Austin, Texas Rebecca Kush, PhD Clinical Data Interchange Standards Consortium, Austin, Texas A Standard Computable Clinical Trial Protocol: The Role of the BRIDG Model Today s clinical trial processes are inefficiently mired in excessive documentation and unnecessary human intervention. The protocol lies at the heart of these operations, making it a prime candidate for the benefits afforded by computer processing. With a vision to develop a standard machine-readable protocol, the Protocol Representation (PR) group, a team within the Clinical Data Interchange Standards Consortium (CDISC), identified a set of elements common to regulated clinical research protocols. These elements are being incorporated into the Biomedical Research Integrated Domain Group (BRIDG) model, an information model of biomedical research initiated by CDISC and collaboratively developed and maintained by CDISC, Health Level Seven (HL7), the National Cancer Institute (NCI), the Food and Drug Administration (FDA), and other stakeholders. The PR group plans to use BRIDG as a pathway to develop a standard structured protocol representation so that protocol information can be repurposed across multiple clinical research documents, databases, and systems from study start-up through reporting and regulatory submissions. The BRIDG model provides a means of knowledge acquisition, a tool for communication, a focus for collaboration, a starting point for application development, a means to standards development and harmonization, and an informatics research platform to support clinical research. Key Words Protocol; Clinical research; BRIDG; Standard; Model Corresponding Address Rebecca Kush, PhD, CDISC, Two Rivers Cove, Austin, TX 78717( rkush@cdisc.org). INTRODUCTION Scientific communication in the biopharmaceutical industry occurs largely through documents, as evidenced by the elaborate document management processes and the volumes of paper that are features of traditional drug development and marketing. Although sophisticated techniques and tools are available for managing these documents, the number of documents that must be managed within any one organization continues to escalate. This increase in complexity limits our abilities to identify, to find, and to retrieve key items of information within and across these documents, as well as our ability to reuse the information effectively and efficiently. Mired in documentation, current clinical research processes are heavily dependent on human intervention, complex data integration steps, and interpretation. The translation and transcription of salient clinical information from these documents into other documents and databases, so that it can be used to drive related internal and external processes, including report generation and knowledge management, is essentially still a manual process. Perhaps the best illustration of this is the complex document that lies at the core of all trials the protocol. As defined by the International Conference on Harmonisation (ICH) of Technical Requirements for Registration of Pharmaceuticals for Human Use, a protocol is a document that describes the objective(s), design, methodology, statistical considerations and organization of a trial. The protocol usually also gives the background and rationale for the trial, but these could be provided in other protocol referenced documents (1). The protocol document is used in clinical research in much the same manner as a blueprint is used in the construction of a building. It is the definitive source for most of the key information about the study throughout its life span, including Institutional Review Board (IRB) evaluation, site setup, recruitment, inventory management, data management, database lock, analysis, reporting, and often beyond to regulatory submissions and postapproval marketing. Currently, this information is buried within a monolithic text document. This creates two problems. First, it limits the ability to assess the Drug Information Journal, Vol. 41, pp , /2007 Printed in the USA. All rights reserved. Copyright 2007 Drug Information Association, Inc.

2 384 CLINICAL RESEARCH STANDARDS Willoughby et al. quality of the protocol. Today s densely written documents are often ambiguous and lack clarity in describing procedures. For example, ensuring that inclusion and exclusion criteria are specified correctly and consistently throughout the protocol is time consuming and resource intensive. Inconsistencies can be found after the protocol has received IRB approval or even after the trial has started, which cause significant delays while the protocol is rewritten, reapproved, and redistributed. A more profound limitation is in the ability to locate and reuse that information, internally and externally, in downstream processes, documents, databases, and reports. For example, we may want to know whether a particular study is a blinded study. At present, one simple approach might be to use a search tool on the text document to search for key words such as study design or blind, but a full-text search usually requires the user to consciously tab through several findings of the word blind until the appropriate usage is identified (eg, This is a double-blind study. ). Furthermore, if synonyms are used (eg, masking for blinding ), such an approach will fail. The surest way is for a qualified individual to read the entire document carefully from the beginning. Since the mid-1990s, ICH guidelines have provided content and format standards for some of the most common clinical research documents submitted to regulatory agencies. Most drug development organizations and major academic medical centers have designed internal templates based on these guidelines and related regulatory requirements. Familiarity with these guidelines, or templates, certainly aids manual navigation, but the information is not sufficiently structured to be usefully interpreted by computer applications in a consistent manner. Furthermore, we often need to find detailed information (eg, degree of blinding) not for just one study but for many and across many documents. For example, a regulatory agency may request that a sponsor provide data from all clinical trials conducted on a drug that included geriatric patients with previous history of heart disease. If the sponsor is not already tracking this, and especially if this drug has been on the market for several years or has multiple approved indications, this information cannot be provided without a substantial amount of manual effort. In addition, recent interest in ensuring transparency of clinical research activities, and providing patients with a means to identify ongoing clinical studies in which they could participate, has encouraged the global development of registries with readily searchable key protocol information. The clinical research and drug development industries would clearly benefit from a computable way to find and repurpose clinical research information. PROTOCOL REPRESENTATION STANDARD INITIAL PROGRESS In 2002, awareness of the aforementioned limitations of protocols prompted leaders from Health Level 7 (HL7), Food and Drug Administration (FDA), and Clinical Data Interchange Standards Consortium (CDISC) to initiate a Protocol Representation (PR) project within the HL7 Regulated Clinical Research Information Management (RCRIM) Technical Committee. This PR group also became a team within CDISC. Membership includes clinical project managers, medical communication specialists, academic scientists, representatives of technology vendors, and statisticians with specific expertise in protocol development for regulated clinical trials. The original scope statement of the group was to identify standard elements of a clinical trial protocol that can be further elucidated and codified to facilitate study design, regulatory compliance, project management, trial conduct and data interchange among consumers and systems. In this statement, standard elements refers to the common component parts or the customary informational units found in most clinical research protocols. Because much of the content for a regulated clinical research protocol has been prescribed by the ICH, the overall hierarchy and the initial set of elements are based on ICH E6 and ICH E3 guidelines (1,2). These are primarily applicable to efficacy and safety trials but can be applied to other types of clinical

3 A Standard Computable Clinical Trial Protocol CLINICAL RESEARCH STANDARDS 385 research as well. Another reference for protocol elements included European Medicines Agency s (EMEA) EudraCT (3) and the CDISC Study Data Tabulation Model (SDTM) Study Summary (4). Later, elements from the CDISC SDTM Trial Design Model (TDM) were also included as well as elements related to statistical analysis based on ICH E9 Guidelines (5). Several other decisions and assumptions helped to lay the foundation for the PR standardization effort; these are outlined in a recently published article by Kush in Next Generation Pharmaceuticals (6). As of 2006, 264 protocol elements had been identified and described. These are posted as a spreadsheet in the Standards area of the CDISC Web site (7). During the development of this spreadsheet, every effort was made to accommodate different types of protocols for clinical research on a global basis. The spreadsheet contains detailed information about each element, including the element name, definition, explanation, and sources for each element and definition. Much of this information was concurrently used to enhance the CDISC glossary (8). Elements are categorized within the spreadsheet by the section heading topics outlined in ICH E6 (Figure 1). An example of the spreadsheet hierarchy, which includes headers, subheaders, and elements, is shown in Figure 2. A STANDARD STRUCTURED PROTOCOL REPRESENTATION The spreadsheet resulting from the initial work to identify and describe standard elements of a protocol is undeniably useful for many purposes, including building consensus around protocol content and meaning and ensuring that all necessary elements would be included in a newly authored protocol. However, the spreadsheet structure is basically flat, lacking information on how each element relates to another, the overall study process, and whether it is a single element or representative of a collection of elements. Following an initial attempt at creating an HL7 model from a subset of the protocol elements, it became clear that the approach and priorities for the PR work needed to be reviewed Protocol Representation Hierarchy Document Type General Information Background Information Trial Objectives and Purpose Trial Design Subject Selection and Withdrawal Subject Participation/Study Design Treatment of Subjects Efficacy Assessments Assessment of Safety Statistics Direct Access to Source Documents Quality Control and Quality Assurance Ethics Data Handling and Record Keeping Financing and Insurance Publication Policy Supplements Protocol Representation Hierarchy Sample: Sections, Subsections, Elements Document Type Clinical Trial Protocol General Information Protocol Identification Protocol Title Protocol Short Title Protocol Identification Number Protocol Contact Information Sponsor Sponsor Status =Section =Subsection =Element FIGURE 1 Protocol representation top-level sections. FIGURE 2 Hierarchy of protocol representation spreadsheet based on ICH E6, E3, and E9. Drug Information Journal

4 386 CLINICAL RESEARCH STANDARDS Willoughby et al. and revised. New directions for this effort included (a) leveraging standards that had matured since the initiation of the PR project, along with the PR elements spreadsheet, particularly the CDISC SDTM TDM and Trial Summary domain [which is identified as a specification in the FDA Final Guidance for esubmissions (9)]; (b) leveraging the domain analysis model that had been initiated to harmonize the CDISC standards and represents protocol as the research plan; and (c) agreeing on an initial set of priority use cases out of the many in which the protocol is involved in clinical research. On the basis of the new direction, the use cases that were agreed to be highest priority were the following: 1. To support the CDISC SDTM (eg, trial design, inclusion/exclusion criteria, planned assessments, interventions, and analyses). 2. To support study tracking databases (eg, EudraCT, clinicaltrials.gov, the protocol/trial tracking aspect of trial registry, or results databases or databases that support project management tools). 3. To support the development of the clinical trial protocol document. The new scope and mission for the PR group thus became To develop a standard structured protocol representation that supports the entire life-cycle of clinical research protocols to achieve semantic interoperability (the exchange of content and meaning) among systems and stakeholders. In this context, the term structured means that the data elements, and the relationships between them, are defined consistently and unambiguously, and are thus computable (ie, amenable to automated processing) and will enable semantic interoperability (exchange of content and meaning). For information from any standard to be acted on by a machine, it must be not only consistent but also compartmentalized, or broken into components or subcomponents that are identified and clearly defined, with the relationships between them likewise clearly defined. This process of assigning identification to information requires people to be unambiguous and precise about meanings and terminology (ie, semantics) and to reach a consensus on meaning where none may have existed before. Any experienced clinical site monitor, investigator, or regulator will quickly confirm that the typical clinical trial protocol document contains many implied meanings and unclear instructions that often lead to misinterpretations, errors, and inconsistencies in trial conduct. This becomes especially complicated when different companies/protocol sponsors have different meanings for the same word. The analysis that must be conducted to create a structured protocol can add value through improved clarity and communication, even if there is no immediate use case for computability of that protocol. In other words, this analysis confers benefits that can be realized even in a paper-based world. In addition to precise, unambiguous definitions for the elements in the PR spreadsheet, there must be a means to capture the information necessary to describe the richness and complexity of the relationships among the protocol elements, that is, a method for representing how these protocol elements relate to, and interact with, other items of information in the clinical trial life cycle. The aforementioned domain analysis model, now named the Biomedical Research Integrated Domain Group (BRIDG) model, fulfills this purpose. THE BRIDG MODEL The BRIDG model is now being approached as a collaborative project led by CDISC, National Cancer Institute (NCI), FDA, and HL7 and includes representatives from these organizations and other stakeholders. The goal is to construct a comprehensive domain analysis model that represents biomedical and clinical research in the context of health care. For the BRIDG model, clinical research is the domain of analysis. The development of the BRIDG model was initiated by CDISC in 2004 as an effort to harmonize existing CDISC models among one another and to harmonize clinical research standards with health care standards. Representatives of the CDISC board developed the initial model, which had the protocol as a major component at

5 A Standard Computable Clinical Trial Protocol CLINICAL RESEARCH STANDARDS 387 its center. The HL7 RCRIM Technical Committee (which was initiated by CDISC, HL7, and FDA) soon adopted this model as its domain analysis model. Using the HL7 Development Framework (HDF) methodology, an HL7 expert led clinical research domain experts through a series of focused modeling and vetting sessions to develop an overarching model intended to represent the entire clinical research domain. The HDF provides a means of eliciting knowledge from subject-matter experts and representing it in a Unified Modeling Language (UML) diagram, using naming conventions and terminology that are familiar to these subject-matter experts. This is the first action in a set of modeling steps necessary for achieving successful harmonization with the HL7 Reference Information Model (RIM), which because of its abstract and complex nature, typically requires HL7 expertise (a highly technical skill set not prevalent within the clinical research community) to navigate and interpret effectively. Concurrent with this CDISC initiative, the NCI instituted a research project called the Cancer Biomedical Informatics Grid (cabig ) a large-scale initiative aimed at creating an interconnected grid of shared data and software tools to support the cancer research community in the delivery of molecular medicine, in the form of innovative therapies, to cancer patients (10). Rather than construct yet another standard for the interchange of protocol information, researchers from the cabig program collaborated with CDISC and HL7 to develop the BRIDG model; CDISC and BRIDG models were considered to be best of breed for the clinical research domain. In November 2004, a team of modelers and clinical protocol experts representing CDISC, HL7, and cabig assembled to begin collaborative work specifically focused on the clinical trial protocol and incorporation of these elements into the existing BRIDG model. Using the initial model developed by CDISC and the 264 protocol elements identified by the PR group as a starting point, the team held a series of meetings over the next six months to construct the preliminary scaffolding of a UML model for protocol-driven clinical research. The process for modeling the protocol meant representing the entire plan for the clinical trial and concentrating on the protocol as a plan for a trial, as opposed to an actually instantiated trial. Placing the 264 protocol elements within the rich representation of the BRIDG model provided a way for subject-matter experts to describe both the what and the how of clinical trials and established a forum for subject-matter experts to clarify the semantics of those protocol elements. The modeling process required domain experts to be rigorous in defining each clinical research information element, in defining the relationships between each element, and in placing the elements and the relationships into the context of specific work processes. This required domain experts to define roles, responsibilities, and the interaction of elements as part of clinical research. For example, knowing that a particular data element is used both for tracking a clinical trial and for registering a clinical trial with external registries would be useful in defining the semantics and the requirements necessary for that element. In addition to the value that the BRIDG modeling effort has had in clarifying the semantics of clinical research and producing a comprehensive and clear representation of clinical research data elements, the process of developing the BRIDG model has served to unify communities within biomedical research, including pharmaceutical development organizations, government bodies, academic medical centers, standards organizations, and related service and technology providers. The HL7 RCRIM Technical Committee has now agreed that the BRIDG model should serve as its domain analysis model for developing all of their standards. CDISC is using the BRIDG model to help harmonize the existing CDISC standards to support the entire life cycle of clinical research. Within the cabig program, the BRIDG model is the starting point in the process of developing interoperable applications that share a common semantic foundation, and harmonization is under way to ensure that these diverse applications and groups Drug Information Journal

6 388 CLINICAL RESEARCH STANDARDS Willoughby et al. can realize the promise of BRIDG and interoperate in a predictable way. The BRIDG model will continue to evolve, even as it is being implemented for specific developed applications. It is an open model, undergoing collaborative development by several key stakeholders. Anyone can contribute to this model by volunteering in one of the stakeholder organizations or by visiting the open-source Web site for the BRIDG project (11). From this Web site, the model can be viewed and downloaded in various formats. Process guidelines for harmonizing existing clinical research information models into BRIDG are in preparation, and an advisory board of representatives from the key stakeholders has been established to ensure that the model remains open, collaborative, and applicable to the stakeholder communities. The model will remain valuable only to the extent that the subject-matter experts within the community contribute to it, interact with it, build on it, and use it. LOOKING FORWARD Although structuring information requires careful analysis of how information should be divided or parsed into smaller units, the goals and intended use for the structured information must drive this analysis to ensure that the resulting representation is useful. The protocol document is used in many activities and serves multiple purposes throughout the life cycle of the clinical trial. The ultimate goal of the structured protocol is to make clinical protocol information backward-referenceable and forwardreusable within and across multiple clinical trial documents, databases, and systems. Three PR priority use cases will enable the development of a standard machine-readable protocol in stages, based on current needs, implementations, and additional use cases. These will both complement and support some of the other BRIDG initiatives that are under way by various stakeholders. The three priorities for furthering the protocol representation work will be described in greater detail; the analogous components of the protocol, which these priorities support, are depicted in Figure 3. PRIORITY 1: SUPPORT THE CDISC STUDY DATA TABULATION MODEL (SDTM) In July 2004, the FDA announced its desire to receive data for New Drug Applications (NDAs) in the SDTM standard format (12). Since then, several NDA submissions have employed the SDTM and more are in preparation. The SDTM is a standard for submission of study data tabulations and also study design information; such a standard for submission of data provides many benefits to regulatory reviewers. Through the use of tools built to read standard SDTM information, reviewers can work with the data more efficiently, with less preparation time. In addition, the FDA and NCI are currently implementing a cross-trial data warehouse (Janus) that will include both information on the study plan/design (ie, protocol information) and the results of those studies. Structured protocol provides the baseline plan so it can be effectively compared against actual data by reviewers. Some of the most important FDA priorities for structured protocol representation include: Trial design Inclusion/exclusion criteria Planned assessments (time and events table) and planned interventions Explicit statistical analysis plan components The PR group is developing structured representations of each of these key portions of the clinical trial plan. The TDM describes a general model for representing treatment arms (represented as horizontal bars in Figure 4), elements within an arm and visits (represented as the cells at the intersections between the horizontal and vertical bars), and is published as part of the SDTM. Inclusion/exclusion criteria will eventually need to be described in a computer-executable manner, which is being pursued. The planned assessments and interventions supports the time and events table, sometimes referred to as the schedule of events, or study calendar, which is found in all clinical trial protocols. Elements of the statistical analysis plan that are typically found in protocols are also included

7 A Standard Computable Clinical Trial Protocol CLINICAL RESEARCH STANDARDS 389 Clinical Trial Registry (CTR); Trial Tracking Trial Design Part 1 Protocol Components for Element Identification, BRIDG Modeling, XML Schema Development Time and Events Table Elements; Trial Design Part 2 Eligibility Criteria Statistical Analysis Plan Elements Machine-Readable Protocol Development and Testing Remaining Protocol Study Report Sections FIGURE 3 Components for developing a machine-readable protocol. Jan 2006 in the PR spreadsheet and are being incorporated into the PR model. When these content models are available, they will be modeled and harmonized into the BRIDG. XML implementations, both extensions of the CDISC Operational Data Model (ODM) and HL7 messages, can then be developed. Work is also under way both to develop interfaces to existing software packages and to develop new applications to support these elements. So another goal of the PR group is to use the PR model as an interface with these tools to facilitate, not only the protocol authoring but also the setup of the trial databases and the data collection (case report) forms based on that protocol. Although the SDTM was designed for submission of data to regulators, it is also used to facilitate data interchange between trial sponsors, partners, and technology or service, providers. If a drug development organization is going to use the SDTM for any reason, it seems logical to increase return on investment by using a protocol standard that is compatible in structure and can be applied throughout the clinical trial life cycle. The recently completed CDISC business case (13), in fact, indicates that the most value from standards applied to clinical research comes when they are implemented in the startup stage of a study, especially in the earliest stages of clinical development. Screen 1st Treatment 1st 2nd Treatment 2nd 3rd Treatment Follow up FIGURE 4 P P P Screen Placebo Screen 5 mg Screen 5 mg Randomization 5 mg Placebo 10 mg 10 mg 10 mg Placebo Follow up Follow up Follow up Example of the arms and epochs in a protocol using the TDM (supporting both the SDTM and protocol representation); courtesy of Diane Wold, TDM Team Leader. Drug Information Journal

8 390 CLINICAL RESEARCH STANDARDS Willoughby et al. PRIORITY 2: SUPPORT CLINICAL TRIAL REGISTRY (STUDY SUMMARY) AND TRACKING DATABASES The next highest priority for developing the structured protocol is the creation of a model to support protocol/trial tracking and the various clinical trial registries, such as the European clinical trials tracking database (EudraCT), clinicaltrials.gov, and the NCI s Physicians Data Query (PDQ ). Such a model would logically encompass a sufficiently representative subset of the eventual structured PR. Building on previous work, the PR spreadsheet of elements, which had already undergone a significant level of critical review, was used to identify a subset of appropriate elements for trial registration or trial tracking. This subset was further developed into a clinical trial registry (CTR) element spreadsheet that has been cross-referenced extensively with code lists for clinicaltrials.gov (14), the World Health Organization (WHO) Registration Data Set v2.2 (15), SDTM Trial Summary Codes (SDTM V3.1.1), PDQ DTD/Schema, and EudraCT. As discussed earlier, because the spreadsheet format lacks the structure required to define the domain fully, these elements were modeled for harmonization into the BRIDG model (Figure 5). Elements that map directly to the WHO Registration Data Set have already been added as an extension to the CDISC Operational Data Model (an XML backbone that carries CRF data and also can carry the SDTM data and metadata), thus enabling electronic WHO registration transmissions today. An HL7 standard is also being considered as an alternate implementation for use in transmissions to multiple trial registries. Inclusion and exclusion criteria are an essential component of trial registries, thus the development of machine-readable eligibility criteria (see priority 1) may actually be viewed as an extension to the CTR work. This is an area of the protocol that is of great interest to many different stakeholders. The stakeholders include registries, those offering ways to match patients to protocols, protocol-authoring software vendors, trial conduct management software vendors, and clinical trials staff eager to reduce the subjectivity of interpreting the eligibility criteria text. A machine-readable standard might also facilitate the conduct of international trials by framing the criteria using international standards and code lists to represent textual eligibility criteria. FIGURE 5 An example of a subset (14 of a total of 65 elements) of the Clinical Trial Registry Model in Preparation of UML modeling for the BRIDG; courtesy of Lakshmi Grama, CTR Team Leader. Protocol Title Protocol Short Title Protocol Identifier Clinical Trials Phase Study Synopsis Participation Type Trial Status Target Study Population Description Target Disease Condition Date of First Enrollment Duration of Subject Participation Targeted Accrual Study Purpose Study Investigation Type Clinical Trials Activities: Study.long Title Clinical Trials Activities: Study.shortTitle Clinical Trials Activities: Study.id Clinical Trials Activities: Study.phaseCode Clinical Trials Activities: Study.description Clinical Trials Activities: Study.multilnstitutionInd Clinical Trials Activities: Study.status Clinical Trials Activities: Study.populationDescription Clinical Trials Activities: Study.targetConditionCode Clinical Trials Activities: Planned Study.TBD Clinical Trials Activities: Planned Study.plannedSubjectParticipationDuration Clinical Trials Activities: Planned Study.plannedSubjectInterventionDuration Clinical Trials Activities: Planned Study.targetAccrualNumber Clinical Trials Activities: Study.intentCode Clinical Trials Activities: Study.TBD

9 A Standard Computable Clinical Trial Protocol CLINICAL RESEARCH STANDARDS 391 PRIORITY 3: SUPPORT THE DEVELOPMENT OF THE CLINICAL TRIAL PROTOCOL DOCUMENT Because this priority has a broad scope, the initial activities toward this use case have focused on working with technology providers who are developing specialized applications based on BRIDG that will use the PR standard. Stakeholders in the BRIDG project understand that standards alone are not useful without the applications that use them. Some individuals have joined the PR group for the purpose of being on the bleeding edge of tool development using the BRIDG. Presentations have been shared at multiple Drug Information Association, HL7, CDISC, Bio IT World, and vendor user group meetings, and several white papers and articles have been published to increase awareness of these projects. A selection of these is included on the CDISC Web site (13) ( This outreach effort has resulted in expressions of interest from several vendors to develop applications using the BRIDG model. Although some are waiting for BRIDG to be refined and expanded to include the areas addressed, many vendors and other groups are actively developing applications using parts of BRIDG and are concurrently enhancing the BRIDG model with their experience. By instantiating the model in software applications, these efforts improve the robustness of the BRIDG model and make tools available to leverage the protocol representation standard in streamlining clinical research. SUMMARY Because all clinical data and conclusions are ultimately rooted in the protocol, the concept of a machine-readable clinical trial protocol has the potential for pervasive impact far beyond the traditional boundaries of that document. Building on ICH content guidelines, a set of elements has been identified as common across the majority of regulated clinical trial protocols and other clinical research protocols. Although this is a necessary step for making information amenable to automated processing, it is insufficient without the added detail provided through conscious and explicit representation of the relationships and interactions among these elements with the rest of the clinical research space. BRIDG has served as a focus within the biomedical research community for a range of activities that will support the goal of a semantically interoperable representation of clinical trials and provides: A method of knowledge acquisition, providing a means of clarifying the semantics of clinical research concepts, and the important relationships and interactions between them. A tool for communication, helping to convey the semantics of clinical trials to people charged with developing applications to support them. A focus for collaboration, bringing together different stakeholders engaged in clinical trials research in an effort to develop and record the shared semantics. A method for developing and harmonizing standards. A starting point for application development for information technology vendors within HL7 and within cabig. The domain analysis model for HL7 RCRIM, which will serve to harmonize all RCRIM standards. An informatics research platform, bringing the expertise of the informatics community to bear on the very concrete problem of understanding and streamlining the administration and workflow of the clinical research process. It is hoped that BRIDG will stimulate new research in informatics, as clinical research workflow becomes better understood. As a key aspect of the BRIDG modeling, the elements and semantics from the PR spreadsheet of common protocol elements have laid the foundation for representing detailed structure of specific use cases for protocol information within BRIDG. The structuring of key pieces of the protocol such as the trial design and the planned time and events table has the potential to support and significantly enhance important use cases elsewhere in the life cycle of a clinical trial from case report development and registration though submission. The BRIDG can serve as a basis for CDISC ODM, HL7 V3 messages, or other implementa- Drug Information Journal

10 392 CLINICAL RESEARCH STANDARDS Willoughby et al. tions. The semantics will be the same (ie, based on a common domain analysis model) so that the information can be readily exchanged and integrated across different tools, applications, and databases, regardless of the way the protocol representation content standard is implemented. Through a combination of structured clinical trial information and standards-based tools, the drug development industry and other areas of biomedical research will reap the benefits that the developers of the structured protocol and BRIDG have been striving to achieve. Acknowledgments We express appreciation to Diane Wold, Joel Hoffman, Greg Anglin, and Art Gertel for editorial assistance with the manuscript. REFERENCES 1. International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use. ICH Harmonised Tripartite Guideline. Guideline for Good Clinical Practice E6; International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use. ICH Harmonised Tripartite Guideline. Structure and Content of Clinical Reports E3; European Clinical Trials Database EudraCT. Available at: Accessed March 15, CDISC Study Data Tabulation Model. Available at: index.html. Accessed March 15, International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use. ICH Harmonised Tripartite Guideline. Structure and Content of Clinical Reports E9; Kush, R. Can clinical trial protocols be standardized? Next Generation Pharm. 2005; Protocol standard protocol representation spreadsheet. Available at: standards/prelements_fulllist2 0-0.pdf. Accessed March 15, CDISC Glossary Group. Clinical research glossary, version 4.0. Appl Clin Trials. 2005;14(special resource issue): FDA Final Guidance (Specifications SDTM). Available at: ersr/studydata-v1.3.pdf. Accessed March 15, cabig. Available at: Accessed March 15, BRIDG. Available at: Accessed March 15, FDA announces standard format that drug sponsors can use to submit human drug clinical trial data. FDA News. July 21, 2004: CDISC Business Case. Available at: Accessed March 15, ClinicalTrials.gov. Available at: Accessed March 15, WHO registration dataset. Available at: Accessed March 15, Lisa Chatterjee has disclosed that she has received financial or material support from Digital Infuzion, Inc. Cara Willoughby has disclosed that she is an employee and stock shareholder of Eli Lilly and Company. Doug Fridsma has disclosed that he receives research support from the National Institutes of Health and the National Cancer Institute. John Speakman has no conflicts to disclose. Rebecca Kush has no conflicts to disclose.