Linked Databases, Health IT, and Informatics Essential for CER and Personalized Medicine

June 2010, Vol 1, No 2 - AACR Annual Meeting


Washington, DC—Health information technology (IT) and informatics are fundamental to all comparative effectiveness research (CER), Amy Abernethy, MD, associate director of the Duke Comprehensive Medical Center and founder and director of the Duke Cancer Care Research Program, told attendees at a session of the American Association of Cancer Research (AACR) 101st Annual Meeting. In particular, patient registries—with their structured inventories of patient data—are a critical platform for CER, assessment of the quality of cancer care, and rapid learning.

Karen Abernathy, MDThe use of information from linked databases can provide real-time information about the benefits and harms of alternative methods to prevent, diagnose, treat, and monitor cancer or improve the delivery of cancer care, Dr Abernethy said at the session, titled Healthcare Reform: The Confluence of Comparative Effectiveness, Health IT and Personalized Medicine.

“Oncology demands science. Right now, it feels like cancer research and our engine for research is being distracted by CER, but when put into context, the importance of CER for cancer emerges,” she said. Although the purpose of CER is to assist consumers, clinicians, and payers in treatment decisions, oncology presents a special challenge because of the wide extent of off-label use of drugs.

“Off-label prescribing is critical in oncology,” Dr Abernethy suggested, since this is “where effective treatment options are often limited, where the prognosis is dire, and there are many potential uses for any given drug and it’s really pretty much impractical to submit an FDA application for every drug-disease combination. As a doctor, biological plausibility and the potential for emerging evidence that would support the use of a drug for a disease means that I am pretty likely to write a script.” That likelihood has real-world implications. “It is estimated that 50% to 75% of all anticancer prescriptions are actually off-label. This is a big deal,” she said.

Legislation dating back to 1993 ensures Medicare reimbursement for off-label use of oncology drugs “provided there is authoritative recommendation for use in the compendium,” Dr Abernethy noted. But an Agency for Healthcare Research and Quality–funded review found that the compendia cited little of the available evidence when recommending an offlabel indication and that the methodological quality of the studies was often poor. In addition, the pace of accumulation of evidence within the context of complex cancer care makes thoughtful evaluation difficult. “How can we be confident that we are providing effective treatments?” she asked. “We need value in cancer care.”

Although clinical trials are the gold standard for effectiveness, they “are typically narrowly focused and can be too short in duration to capture certain important outcomes, such as overall survival,” Dr Abernethy said. “CER focuses on populations but may not provide insight for the best care of individuals. In oncology, we take care of individuals.” Because “no single study design answers all the questions,” CER must draw from a broad methods’ portfolio, she said.

The Institute of Medicine has identified cancer as the prototype for a “rapid learning healthcare approach.” With this, “data collected during routine patient care is fed into an evergrowing, coordinated set of data – bases,” Dr Abernethy outlined. The system learns by routinely analyzing captured information, integrating it, generating new evidence, and constantly tailoring new insights into personalized care.

“Critical to this process is linked information,” she said. “In rapid learning, the care of this individual patient is informed by all the people who came before her, and her care is reinvested into the data stream to contribute to the care of similar individuals in the future.”

A first step in developing rapid learning is to leverage information in existing databases, such as patient registries. “We want to promote a method of CER that conserves clinical trial resources for important, prioritized clinical questions. We want to use databases and data mining to explore real-world effectiveness, transfer clinical trials data to populations, and highlight compelling new hypotheses to subsequent study,” Dr Abernethy said. “Rapid learning healthcare will contribute continuous real-time knowledge to understand what works for whom and when.”

Build It and They Will Share

An estimated $48 billion will be spent over the next 4 years on largescale activities to create a national infrastructure for healthcare IT, Ken Buetow, PhD, National Cancer Institute (NCI) associate director for Biomedical Informatics and Informa – tion Technology, told the session. The efforts “dwarf anything we would be doing in the NCI related to health information technology,” he said. “However, it is important to note that the current strategies largely are divorced and disconnected from the broader research framework” and have little consideration for how “to bring all this information together and to support all these invaluable uses of information that will be in the system.”

Dr Buetow continued, saying “the primary infrastructure that is being constructed is around primary care delivery in the context of individual physician offices, with very little knowledge or interest in how…to go about interconnecting all of this information so that the whole can be more than the sum of the parts.” Dr Buetow sees a unique opportunity for cancer researchers and providers to lead in healthcare IT. “We want to see to it that the cancer community has the capacity to be wired into this broader national framework, but more importantly, that we will be an example for how we can interconnect and use this information in powerful and meaningful ways.”

Led by the American Society of Clinical Oncology, NCI has been working on the creation of “smart” electronic health records (EHRs). The records use a common structural organization so that provider data “is in the same form as data we use in clinical trials research,” Dr Buetow pointed out. Such standards are important to facilitate information sharing by all users. Com munity physicians are “hungry in terms of what we can give them” in the best science tools and infrastructure. “They really want to be part of that framework” and have access to state-of-theart treatments, protocols, and tools to see how patients are performing.

The EHRs are part of the Institute's Cancer Biomedical Informatics Grid and will enable collection of cancer diagnostic and staging information, treatment plans, and patient outcomes in the care setting. Clinical information can be fed into an electronic clinical report form as well as personal health records for use by consumers. Web-based tools will allow consumers and patients to record their personal information and treatment responses.

Data on patient encounters can be fed into a patient outcomes resource that can help physicians individualize care, aid patients in understanding their options, and enable researchers to query data. In order to protect confidentiality, information would be stripped of common identifiers. In – formation collection “could also be done in a manner in which people are consciously providing data,” he said. “The patient can be incredibly important to drive the next generation of studies.”