Life Science Explorers – Uncovering R&D Intelligence

by Jason Burke

Within the information technology community, the volume of corporate information is doubling every 12 to 18 months, according to estimates. This is especially true in life sciences. In drug development alone, companies are facing growing volumes of data from traditional case report forms, electronic data capture systems, medical records, genomics, central laboratory facilities, medical devices, business partnerships, and patient diaries, just to name a few.

To cope with this growing data deluge, companies are turning to new classes of software focused at managing the life cycle of data. Enterprise information management (EIM), a term cited by Gartner, typifies this new focus on understanding the process of managing data over time, as well as the relationship between that data and other information available within the company.

But the challenge facing life sciences firms is not just developing an EIM strategy. In the day-to-day course of business, data is used many different ways. Transformations are conducted, views are created, analyses are designed and executed, and important medical decisions are made. So EIM for life sciences is not just about aggregating and integrating the data – it should also address how the information is processed and used.

Charting the course
Despite the thorough methodological and empirical underpinnings that define scientific research as a discipline, science is also an exploration. Like the early explorers of America, today’s researchers use increasingly more detailed “maps” to guide their direction – more complex research designs encompassing more detailed research data. And they need a way to document and retrace their steps through the unpredictable terrain of drug discovery and development. They need to be able to quickly return to particular locations and know what decisions were made using what information.

Software is playing a definitive role in helping to address these needs. Consider:

• If you have data coming from multiple sources to multiple people, how do you know where the data traveled and what happened to it in transit? By using software to standardize and automate the logging and processing of inbound data, you can reduce errors and problems with the timeliness of the information. The “steps taken” on the path are consistent, documented, enforced and compliant. You might use the software as a sort of “electronic lock box,” ensuring that you always have an unadulterated copy of the information to return to if needed. Part of the processing might even align the data more closely with your corporate standards by providing additional metadata to the incoming data – making the new information more useful and valuable later.

• If your data is constantly being updated, how do you know who looked at what data to make what decisions? Audit trails could allow you to view a copy of the data and activity at a particular point in time, allowing other explorers to follow your path later. An even better solution would allow you to standardize who sees what information (regardless of where the information came from), and how they see it – online viewing, reports, etc. If the research process includes people from different organizations (as it almost always does), software can provide a secure electronic space for those people to collaborate in the research design, data handling and decision-making processes.

• As your statistical analysis evolves to answer additional questions raised by the research, how can you track the evolution of that analytic process? Using a statistical computing environment that includes version control of statistical algorithms, input data and output results would allow you to see part of the evolution of thought associated with the research findings even as the data was changing. An ideal solution also would allow users without technical or statistical training to have access to the information in a controlled and user-friendly manner so that the evolution of statistical thought and clinical thought could be tracked in parallel.

• As the volume of scientific information increases, how can you make better reuse of the research you have already conducted? As the scenario above describes, existing data management practices could be extended by providing standards, automation and additional metadata that would allow other researchers to glean enough insight into your research repositories to construct their own analytic models. Note that the intention here is not to create one place to store all of the data; rather, the focus is on defining a methodology to enhance your processing of the information regardless of where it is stored so that you can derive more value from the research over time. One can see these concepts being explored in other areas of scientific technology today such as federated data stores and the Semantic Web.

These views of information management – data management and process management applied to scientific data – begin to move away from the traditional models for how statistical analyses and clinical insights are obtained. This deeper level of insight, which we might term “R&D intelligence,” focuses not just on understanding the research design and outcomes. Rather, scientific intelligence offers insight into the process by which scientific insight is being achieved, and offers capabilities for improving future scientific explorations. It enables organizations to include “time” as a component to understanding the progress of their scientific work. And though many see this evolution as relating to the activities at the end of a clinical research study, exactly the same approach can be used to provide the analytic underpinnings for adaptive trial design much earlier in the process.

Standards help guide the way
Within clinical research, the development and adoption of industry standards, such as those by the Clinical Data Interchange Standards Consortium (CDISC), have greatly helped organizations move toward these capabilities. The use of data standards is pivotal to being able to provide the needed data aggregation described above. And as organizations explore how best to implement these data standards, there is a growing recognition that other types of interoperability, beyond data standards, will be needed as well. Over the past two years, a dramatic increase in interest and participation can be seen along a wide array of standards-related initiatives – HL7, caBIG, BRIDG, SAFE, Janus – all of which represent components of the longer-term goal of interoperable technology architectures.

But companies can move toward R&D intelligence today. SAS provides the software many life sciences organizations use to design and execute analyses of all types – scientific findings, sales and marketing data, manufacturing quality, and many others. SAS Drug Development is a solution specifically designed with the ideas of R&D intelligence and EIM in mind. The system helps organizations improve their information management by exploiting CDISC standards within a compliant statistical computing environment. Daiichi Medical Research is just one of a rapidly growing customer base that has already looked to SAS Drug Development to improve clinical information integration, transformation, analysis and decision making.

So where might the industry move in the future using EIM? Since most of the effort in drug development today revolves around creating a series of predefined deliverables – a final study report, new drug application, federal submission, etc. – it is easy to imagine managing those deliverables as R&D intelligence objects as well. For example, a solution might leverage industry standards like CDISC to help automate the design, testing, and management of research models before the launch of a research protocol. Statistical programs and their associated outputs could be associated with their corresponding sections in a final study report, enabling users to manage sections collectively rather than as individual files (e.g., “update all tables in this section”). Or a solution might be able to automatically manage and report the life cycle of each variable in the study – where the variable’s data came from, how it was processed and where it occurs in the final report.

Though no vendor has committed to delivering any of these future ideas, it is certain that the idea of gaining deeper insight and control over expanding scientific information sources will continue. As the lines between drug discovery, clinical research and drug surveillance continue to blur, we should expect that the goal of understanding the scientific process as it actually evolves will be more important. Successful clinicians and statisticians will be able to describe in much greater detail what they have seen, what courses they then chose to pursue, and what new explorations may lie ahead.