Developing Standardized Data: Connecting the Silos

We live in an electronic age where information is captured every second, and we have all learned to accept standards as a regular part of our lives. For example, the letters that we send through the postal service would likely never reach their final destination without a standardized format for addresses. In addition, when fire engines come from other municipalities to put out a fire in our neighborhood, we aren’t concerned with whether or not their trucks will be able to hook up to the local fire hydrants, because all connections are standard. Contrast this with the Great Baltimore Fire of 1904 where thousands of fire fighters from the surrounding cities and states were unable to assist as the fire raged for 30 hours, simply because their fire hoses were not compatible with the Baltimore hydrant connections. ¹ At present, we are witnessing an ever-increasing and soon to be overwhelming stream of data in our regulatory review pathways from electronic health records and clinical trials. Regulatory stakeholders, much like the fire fighters from 1904, need to develop those standard connections that will allow us to capture and fully utilize the wealth of available data in order to assist in addressing our most challenging health problems.

While information capture typically calls to mind consumer-targeted data collection, the medical and regulatory review settings also acquire massive amounts of data. Patients may be continuously monitored, and those data may be collected, stored, and used immediately for clinical decision making or used later for analysis of various metrics such as outcomes. Ideally, a patient’s medical information could be easily and automatically available to health care providers, regardless of the point of care; however, data from electronic health records captured in one medical facility are often not easily transferred to another. This clearly presents a hurdle in leveraging all available patient data in the health care setting. A similar hurdle exists for the data gathered from clinical trials where much of the data exist in “silos” because they are not collected consistently and may need to be converted from one form to another before submission to the regulatory agencies. Even essential data variables may be formatted differently between trials for a single product, across drug trials within a class, and for trials across drug classes. If these “silos” were connected, regulatory stakeholders would be empowered to more efficiently review data and ask more probing questions.

One of the simplest examples of data standards challenges that regulatory agencies encounter is the representation of males versus females in clinical trials. The data may be presented in various ways such as “male and female,” “M and F,” “1 and 2,” or “0 and 1.” As one can imagine, inconsistencies of this sort can cause a host of problems when trying to group data from multiple clinical studies together and make assessments within and across therapeutic areas nearly impossible. Early efforts are under way at FDA to address these emerging data standardization needs.

Several initiatives launched within the FDA’s Center for Drug Evaluation and Research (CDER) are aimed at the development of data standards and management of clinical data across specific therapeutic areas. For example, the Office of Clinical Pharmacology in the Office of Translational Sciences in CDER has developed an infrastructure for pharmacometric knowledge management in a subset of disease areas. This infrastructure includes data standards development, queryable databases, libraries of modeling tools, and archives of analysis results. ² One example of the success of this approach is the Antiviral Information Management System (AIMS), a database of Hepatitis C virus (HCV) data in a structured and standardized format. This database has assisted in improving dose selection during early phase drug development meetings and was essential to the regulatory analysis to two recently approved groundbreaking direct-acting antiviral therapies for HCV disease. ³

CDER has also developed a comprehensive data standards program to identify and prioritize data standards needs and to implement good practices for standards development. ⁴ This program has dedicated space on the FDA website to communicate advances in data standards research and regulatory recommendations for data standards with external stakeholders. CDER has engaged with standards development organizations as well as industry and other stakeholders to develop standards where current specifications are not available or are not adequate.

In addition to CDER’s work with external stakeholders, the FDA is embarking on a unique approach that is poised to provide tangible results in the data standards arena by collaborating with a number of stakeholders. At the March 2012 FDA/Pharmaceutical Users Software Exchange (PhUSE) Annual Computational Science Symposium, working groups initiated discussions on validating data, improving data quality, standardizing data within the site selection process, exploring the challenges of integrating and converting data across studies, identifying standards implementation issues with the Clinical Data Interchange Standards Consortium (CDISC) data models, developing standard scripts for analysis and programming, and creating the nonclinical road map and its impact on implementation. These working groups will continue to meet through the year. More information on these efforts can be found on the PhUSE website. ⁵

The Critical Path Initiative, introduced by FDA in 2004, has encouraged industry, academia, and government agencies to develop public–private partnerships (consortia) in order to collaborate and share information, technology, and expertise to bridge the gap between scientific discoveries and their translation into innovative medical therapies. Two examples of partnerships that are incorporating data standards into their efforts are the Coalition Against Major Diseases (CAMD) consortium and the Analgesic Clinical Trial Translations Innovations, Opportunities, and Networks (ACTTION) public–private partnership. CAMD, a consortium convened through the Critical Path Institute, worked with CDISC to develop a user guide for standard data elements for Alzheimer’s disease (AD) and utilized these standards to create a database of AD studies including data from more than 4000 patients. ⁶ The ACTTION public-private partnership is developing a standardized analgesic database platform to analyze previous trial data in order to facilitate the development of enhanced pain management interventions. ⁷ The groundbreaking efforts of these partnerships have encouraged other consortia to take steps toward data standardization.

In the not-too-distant future, our society will be overwhelmed by data, particularly in the medical field. Early development of standardized representations of critical data elements will prevent the formation of “silos” and allow us to connect the data from multiple studies in order to facilitate drug development and optimize the detection of safety signals. Without standardized approaches for the representation of data, regulatory agencies are left like the Baltimore firefighters in 1904, unable to fully access those resources that are essential for optimally performing our tasks.