3 November 2011. Try combining random pages from a novel and then figuring out the plot. That is similar to what regulatory agencies are up against when they analyze results from drug applications for Alzheimer’s disease (AD), with each company using its own system to collect and organize patient data. (see ARF related news story). Although the new standard was developed primarily to improve the AD drug approval process, if it is widely adopted and replaces other standards currently in use, it will help anyone trying to pool clinical data from disparate studies and databases to conduct large-scale analyses and comparisons.
Since 2004, the Center for Drug Evaluation and Research (CDER) of the U.S. Food and Drug Administration (FDA) has strongly encouraged companies to use the Study Data Tabulation Model (SDTM) standard developed by Clinical Data Interchange Standards Consortium (CDISC), a not-for-profit global organization whose mission is to develop standards for medical research and healthcare. “The volume and complexity of non-standardized drug-related information submitted to CDER for regulatory review is creating significant challenges to the Center’s ability to efficiently perform its critical public health mission. The lack of standardized data affects CDER’s review processes by curtailing a reviewer’s ability to perform integral tasks such as rapid acquisition, analysis, and reporting of regulatory data,” wrote Yolanda Fultz-Morris, a spokeswoman for CDER.
The original SDTM, which went through several iterations, provides a base vocabulary primarily for reporting safety data (i.e., adverse events, concomitant medications, exposure, vital signs) across all diseases; however, it leaves out measurements unique to each disease. This core guide needed to be extended with variables and terms for efficacy data that are related to AD-specific clinical endpoints. To do that, CDISC teamed up with the Coalition Against Major Diseases (CAMD). CAMD is the AD initiative of the Tucson, Arizona-based Critical Path Institute (C-Path), a collaboration of pharmaceutical companies and independent and regulatory scientists that works to make the drug approval process more efficient (see ARF related news story).
The impetus for establishing AD standards came once CAMD began to create a database of pooled results from the placebo arms of past AD trials conducted by member companies (see ARF related news story). “CAMD came to us and said ‘We are compiling an Alzheimer’s database from trials; can you help us create a common standard that will enable us to map and better analyze the data?’” recalled Bron Kisler at CDISC. In essence, "mapping" data to a standard is computing lingo for connecting the dots; it involves relabeling individual data items according to a set vocabulary and then reordering them into a predetermined structure so that every corresponding data item from different trials is called the same thing in the standard language. For example, researchers implement the AD Assessment Scale Cognitive Test Battery (or ADAS-cog) in different ways in different trials, asking patients questions in their own distinct order and labeling answers with unique terms. The new AD standard aligns the ADAS-cog tests to account for those distinctions, so that the common data obtained will be recorded in precisely the same way every time. The same concept applies to other psychometric tests, and genotype and biomarker tests. The new standard will also include guidelines for hippocampal measurements based on magnetic resonance imaging (MRI) and MRI protocols, but those are still in draft form, according to Jon Neville at CAMD.
The AD standard is part of a bigger effort by CDISC to develop disease-specific data standards. The FDA enumerated 55 diseases in need of a standard language, with AD listed in the first of three priority groups (see FDA table of Priority Disease/Domain Areas for Data Standardization). A data standard for Parkinson’s disease should be released next year.
Today, the CAMD database, called C-Path Online Data Repository (CODR), contains detailed clinical information from 15 industry-sponsored trials from Abbott, AstraZeneca, Eisai, GlaxoSmithKline, Johnson & Johnson, Pfizer, and Sanofi-aventis, and one academic trial, the homocysteine study, from the Alzheimer’s Disease Cooperative Study (ADCS). Altogether, the database includes details from more than 4,900 patients with mild cognitive impairment (MCI) and AD, all mapped to the new AD CDISC standard. Any researcher whose requests are accessed through CAMD and get approved can mine these data.
Companies have an incentive to adopt the new standard moving forward, just as many adopted earlier versions of the CDISC standard. “Since 2004, the number of submissions to CDER with CDISC SDTM datasets has increased,” wrote Fultz-Morris. “The development of therapeutic area data standards for the CDISC SDTM is a critical component toward ensuring that FDA receives high-quality standardized data to allow more efficient and effective review of new drug applications.” In 2010 and 2011, about half of all clinical trial datasets abided by CDISC SDTM (see FDA report).
Another reason the FDA would like data in the same standard is that, in addition to evaluating investigational new drug (IND) and new drug applications (NDA) from individual companies, CDER pools and analyzes study results from different companies and academic researchers to evaluate the long-term safety and efficacy of prescription and over-the-counter drugs. Having all the data in a standard format simplifies that process. “The FDA wants to spend less time organizing, accessing, and analyzing data, so that they can focus more time on the actual review,” said Kisler.
The ideal time to implement the CDISC standard is right at the start of a clinical trial. “You create case report forms that incorporate standard data elements so that, as data accumulate in the repository, they are automatically annotated in the CDISC format,” said Enrique Aviles of C-Path. “Collecting the data this way requires no more effort than doing it any other way.” The FDA concurs: “CDER strongly encourages IND sponsors and NDA applicants to consider the implementation and use of data standards for the submission of applications. Such implementation should occur as early as possible in the product development life cycle, so that data standards are accounted for in the design, conduct, and analysis of studies. The purpose of the CDISC SDTM Alzheimer's disease/Mild Cognitive Impairment User Guide is to foster more consistent implementation and use of the SDTM for Alzheimer’s disease studies,” wrote Fultz-Morris.
Sounds simple enough, but there is a hitch. Some drug testing programs have been ongoing for several years, so switching to a new standard midway is not always feasible. “Usually the databases for studies are set up prior to any data being generated in patients, so a standard is selected for the database prior to enrolling patients,” said Brian Corrigan of Pfizer. “Many programs, which may have started more than a decade ago, will adhere to a standard that was used initially so that the data from a program is all in one format when it is collected and databased, and subsequently compiled for submission,” Corrigan added. To have data all in the new format means that companies will have to go back and map those older studies to the new standard—a labor-intensive process. “We can’t just flip a switch and turn data into a different format. When we or others change the way we handle data, there is an upfront cost, but in the end it can help efficiency,” said Eric Siemers of Eli Lilly and Company.
Most companies will now be asking their statisticians and database curators to analyze the new AD standard to determine how it differs from the one they have been using, and when and how to make the switch, if at all. “For companies that have been using the CDISC standard, this new one is not that huge a jump, but the investment required may vary from program to program,” said Corrigan. “For a program that is just a week away from regulatory filing, it may not be worth it to map all the data over to the new standard, but for something that is just getting off the ground, it would make more sense.” Pfizer was one of the companies closely involved in the development of the new standard.
Many academic programs have not been using the CDISC standard but developed their own. The Dominantly Inherited Alzheimer Network (DIAN) collects clinical and biomarker data on families affected by early-onset autosomal-dominant AD, and is now gearing up to run clinical trials in this group. DIAN uses both the standards of the National Alzheimer's Coordinating Center’s Uniform Data Set and of the Alzheimer's Disease Neuroimaging Initiative (ADNI), a program within ADCS (see ARF ADNI story). These standards evolved separately from the CDCIS standard. For longitudinal studies, switching narratives may present challenges, according to Virginia D. Buckles at Washington University in St. Louis Missouri, where DIAN is headquartered. “If we change our testing protocol and data collection from what we collected three years ago, we may not be able to use data we obtained at baseline,” she said. Her group has not yet had time to evaluate the new CAMD-CDISC standard, but she agreed in principle with the effort to standardize information. “If everyone uses the same measures, the science will definitely go much faster,” she added.
For ADNI, which relies heavily on imaging measurements, the new standard may not provide sufficient guidance in that area. According to Aviles, CAMD has been meeting with ADNI representatives to discuss how to map ADNI data to the new standard. There are, however, currently no plans for ADNI and ADCS to switch to the CAMD-CDISC standard, according to Paul Aisen at the University of California, San Diego, who heads ADCS. “ADCS trials are not regulatory trials, so whether we will adopt the standard is not clear at this point,” said Neil Buckholtz of the National Institute on Aging in Bethesda, MD, which funds ADCS. “But I think the new standard is very important for the purpose of comparing various trials. It is not easy to do that right now because people use unique fields to identify different data items; even a data field as simple as gender can be captured in many different ways,” he said.
Another academic group that is planning to run clinical trials is the Alzheimer's Prevention Initiative (API) at the Banner Alzheimer’s Institute in Phoenix, Arizona (see ARF conference series). Both the API and ADCS are planning trials in preclinical AD (see ARF related news story and ARF news story). Pierre Tariot, who co-directs the API at Banner, said that, although the initiative plans to go through the regulatory process, just like a company would, they have not yet decided what outcome measures to use and could not comment on which data standard they will choose. Like many others, however, he supports the CAMD-CDISC initiative to provide one global standard for AD research. “In general, standardization is going to be helpful in that you can compare data and harmonize key outcomes,” he said.—Laura Bonetta.