A Proposed Clinical Trial Model: Analyzing the CT Process

Over the past 15 years, methodology and ethics have become important issues in the design and conduct of multicenter clinical trials. Nowadays, clinical research embraces very different disciplines, such as information technology, regulatory affairs, quality assurance, and economics, and it is associated with a complex regulatory framework. The recent EU Clinical Trials Directive has added even more to this complex framework. All of this makes clinical research a more complicated process, and this translates into a difficult planning phase, increased trial time, increased costs, and additional energy.^1-3

Generally, the start up of a clinical trial requires a new design, and once the trial is completed the entire process used in the design is discarded. There is no system currently available that can manage the activities of clinical trials based on the interoperability of each single task. A detailed model of the entire CT process would allow for controlling the process in all its phases, in particular:

• Writing the protocol coherently and consistently so that it supports the collaborative work of the writing committee members

• Facilitating and coordinating the setup of the trial in the various participating centers

• Supporting the critical phases of the process, such as the eligibility, drug supply, severe adverse events (SAE), etc.

• Monitoring the setup, particularly the quality of clinical data and timely collection of case report forms (CRFs)

• Supporting the evaluation and diffusion of clinical trial results.

There appears to be a strong demand for integrating research and systems in a way that supports the whole clinical trial process.⁴ Such an approach can foster interoperability between CT software applications, making system reengineering and maintenance easier. For this reason, we propose a clinical trial model that provides a more detailed analysis of the entire CT process to better understand the number of procedures carried out during the trial.⁵ Our approach is based on a holistic view of CTs. We are developing a comprehensive conceptual model of the CT process independent of the possible information and communication technology (ICT) solutions applied to the process automation.

Figure 1. Process types in the clinical trial process.

Aim of the project

In order to develop a standard model and a common terminology, GIMEMA (The Italian Group for Haematological Diseases of Adults) together with the Italian CNR (National Research Council) have developed the first version of a detailed model of the entire CT process that focuses attention on:

• activities that compose the CT process

• rules that govern its functioning

• data needed to synchronize CT activities carried out by the different actors involved (e.g., sponsor, investigators, data managers).

From the analysis of this model, it will be possible to:

• verify the pillars of the trial and, thus, identify the different types of support that ICT can offer in this field (e.g., applications for communication between and within centers, CRF and serious adverse event (SAE) management, and tools for execution monitoring)

• help and support the writing of a consistent protocol that describes the entire process in a detailed, comprehensive, and coherent way for all actors involved in the performance of the trial

• determine which parts of the process are independent from the type of trial and which ones vary, thus helping to achieve a standardized protocol text.

The clinical trial model

We propose to model trials as a highly organized process that is composed of subprocesses.⁵ We have identified three main subprocesses. Each subprocess has different objectives and is enacted in different environments, carried out by its own agents, supplied with needed resources, and governed by specific rules. The three main subprocesses are:

• The clinical trial management process (CTMP), which includes activities related to the setup, coordination, and monitoring of the CT participating centers, and the final evaluation of trial results. This process is carried out by the sponsor.

• The statistical units management process (SUMP), which includes activities related to the management of the trial in the participating centers. In our approach to trials, a statistical unit indicates an instance of an enrolled patient. The process described in the protocol is executed in each center and performed on each enrolled patient.

• The patient health care delivery management process (PM), which relates to the diagnostic and therapeutic activities necessary to treat an enrolled patient, following the instructions defined in the protocol. This process is carried out within the clinical ward (CW).

Figure 1 describes the relationship between the three process types. As soon as the clinical protocol is developed by the writing committee and approved by the scientific and ethics committees, the sponsor opens participating centers to enrollment. This is why in the figure the CTMP triggers the SUMP for each participating center. In each center environment, for each enrolled patient a PM process instance is created. This means that at a certain point in the SUMP there are as many PM instances as the number of enrolled patients in treatment. For the SUMP, these instances represent statistical units that will be evaluated during the analysis of the clinical trial results.

Figure 2. Data flow of the clinical trial.

Information flow

In our approach, modeling a process means to identify the main participants involved, the needs of participants with regard to what procedures need to be carried out, the objects used (resources) and delivered (results) by the process, and the rules that constrain the execution process. Generally, two processes may be synchronized on certain events or objects: One process (the consumer) will not start because it has to wait for an event or object produced by the other process (the producer). Some of the subprocesses are simultaneous and their interaction may be triggered when specific clauses/conditions are produced because of the occurrence of certain information/documents or specific events.

Figure 2 shows the information flow between the trial processes and specifies the different types of information exchanged (e.g., data, resources, commands/directives, workflows/process specifications, etc.). This information is contained in at least one of the following three highly structured documents:

• The CT master file

• The CRF database

• The patient's Health Care Record (HCR).

The CT master file. This file contains the information necessary to set up and perform the entire CT. The whole CT process can be divided into three sequential macro phases: preparatory, execution, and evaluation. In the preparatory phase (which is represented by the CTMP in Figure 1), the focus is on planning and specifying what needs to be done in the following phases. In the first phase, one of the main tasks is the development of a CT master file. The CT master file contains, among other documents, the protocol (i.e., the document that describes the trial's objective(s), design, methodology, statistical considerations, and organization) and the CRF templates (i.e., a document designed to record all of the protocol required information to be reported to the sponsor on each trial subject). After defining the contents of the CT master file, two processes can start: database creation and writing of the operating procedures manual. The start-up experimentation process can begin only when the previous three processes are concluded.

Groups Behind the CT Model

The CRF. The CRF database gathers the CRF from all patients. A compiled single CRF form describes the patient's health status in each well-defined phase of the CT execution; the whole CRF also reports the changes in the patient's health status occurring during the CT. Analyzing the type and number of compiled CRF forms makes it possible to monitor the CT execution.

The patient's HCR. The HCR is the official repository of the clinical information related to a patient in a given CW. It contains both the information on the therapeutic approach and changes in the patient's health status. Data needed to fill in the CRF are extracted from the HCR, and the quality of the CRF data is verified through the HCR.

Application of the model

The model is defined by a formal language (UML), which describes the process from three different points of view: functional, structural, and behavioral.⁶ Each view highlights a specific aspect of the trial (e.g., the flow of activities, the data structures, and the interaction between the various activities involved in the process).

We used the model to outline areas where ICT can improve and optimize the whole process. In particular, the model has been used to:

• Design an information system, which helps the protocol writing and guarantees its completeness and coherence as well as the extraction of information necessary to automate single activities of the trial

• Formalize the rules that govern the exchange of data in the CT process, which enables interoperability between the various ICT systems managing the single activities of the process

• Define a procedure that guarantees the quality of the CT process related to all its activities.

The protocol is a planning document for the whole CT, where all activities of the process are taken into account. The functional view of our model is an excellent guide to protocol development. This view also includes data necessary to perform each single activity in the CT process, which has allowed us to define a structural framework that guides the text editing and controls its consistency. The framework has been specified by an XML scheme, which facilitates search functionalities within a database of protocols, re-use of single parts contained in the text, and interoperability with other systems.

We implemented a system, WITH (Write on Internet clinical Trials in Haematology) that helps the CT writing committee monitor the different phases of the writing process, re-use parts/sections already written belonging to previously developed protocols, and promote the text coherence obtained through the management of XML tags as well as a related data dictionary.

Another important aspect of the process, which has been emphasized by the model, is the exchange of information between various activities. The structural view of the model allowed us to define an XML format of the exchanged messages, guaranteeing the interoperability between the parts of the system. In our case, the achievement of interoperability also makes it possible to develop and test autonomously single modules of the system, which can be gradually implemented and easily updated to standards such as HL7 (Health Level 7) and the CDISC Operational Data Model.^8,9 The use of standards guarantees the provision of seamless and automatic connections from one software to another, regardless of platforms, applications, or programming languages.

Since our model provides a reference framework for all CTs, it can help achieve the uniformity of protocol quality evaluation. Moreover, having a controlled system that governs all single steps regarding the conduction of a given trial can guarantee that the main regulatory issues are respected. As a result, a standard good practice can be developed that guides the stakeholders (physicians, investigators, sponsors, etc.) during the CT process.

Conclusion

The process model we are studying is the starting point for defining standards related to the use of ICT in CTs. In fact, this model identifies how data (messages) between the various activities should be exchanged. These messages need to be standardized in order to achieve semantic interoperability for the automation of the entire clinical research process. At a higher level, the CT model enables organizational interoperability by defining a virtual community of organizations playing roles as CT coordinator, participant center, etc.

P. Fazi, MD, L. Collada Ali, *project manager, and M. Vignetti, MD, are with the The Italian Group for Haematological Diseases of Adults (GIMEMA), Dip. di Biotechnologie Cellurali ed Ematologia, Universita degli Studi di Roma "La Sapienza," Via Benevento, 6, 00161 Rome, Italy, +(39) 06 441 639 828, fax +(39) 06 440 2516 639 828, email: collada_ali@bce.uniroma1.it.

D. Luzi is a researcher and F.L. Ricci is a researcher with The Italian National Research Council (IRPPS-CNR), Rome, Italy.

L.D. Serbanati is a professor with UPB - "Politehnica" (University of Bucharest), Bucharest, Romania.

References

1. D.L. Rubin, J.H. Gennari, S. Srinivas et al., "Tool Support for Authoring Eligibility Criteria for Cancer Trials," AMIA Symposium 1999, available at http://www.amia.org/pubs/symposia/D005609.PDF.

2. J. Unutzer, Y. Choi, I.A. Cook, S. Oishi, "Clinical Computing: A Web-based Data Management System to Improve Care for Depression in a Multicenter Clinical Trial," Psychiatric Services (June 2002).

3. Verny and I. Klingmann, "A 6-Month Process for Planning Multinational Clinical Trials," Applied Clinical Trials, 58–61 (February 2003).

4. Standardizing Information Flow in Clinical Trial Protocols. Panel organizer: John Gennari, available at http://www.amia.org/pubs/symposia/D200541.PDF (2000).

5. L. Collada Ali, P. Fazi, D. Luzi, F.L. Ricci, L.D. Serbanati, M. Vignetti, "Toward a Model of Clinical Trials," in Biological and Medical Data Analysis, 5th International Symposium, ISBMDA 2004, Barcelona, Spain, November 2004, Lecture Notes in Computer Science (Berlin: Springer-Verlag, 2004).

6. J. Rumbaugh, I. Jacobson, G. Booch, The Unified Modeling Language Reference Manual (Addison Wesley, 1999).

7. P. Fazi, D. Luzi, M. Manco, F.L. Ricci, G. Toffoli, M. Vignetti, "WITH: A System to Write Clinical Trials Using XML and RDBMS," AMIA Symposium 2002, 240–244.

8. http://www.hl7.ca/.

9. D. Iberson-Hurst, "The CDISC Operational Data Model: Ready to Roll," Applied Clinical Trials (July 2004).