"Source" or "esource" could be on the grapevine this year, as sponsors scramble to determine if their systems comply with a new draft guidance from the FDA. It was issued in December, 2010. (You can download it here.)
There is nothing controversial about the guidance document. This is good. In an ideal world, people being regulated and the people doing the regulation should be on the same page. They should be working with the same concepts, the same basic understanding of the territory. Generally, in biomedical research in the U.S., the regulated and the regulators are in sync.
But in subtle respects, the new guidance shows new thinking at the FDA. The agency is clearly redefining one of the most basic concepts in biomedical research. That could nudge the industry toward systems or practices it is not now using. It could also sow confusion.
In the draft guidance, the FDA is using a new definition of "electronic case report form (eCRF)." It is above a journalist's level of brainpower to declare which definition of eCRF should prevail. But there is a difference of terminology.
The industry defines an eCRF as a collection of a few data elements, corresponding to a computer screen of data. It amounts to what might have been collected on a single piece of paper. This definition of eCRF is inherently singular, unitary and backward-compatible with a common mental construct of a paper form.
For the FDA, an eCRF is inherently about multiple forms, collections of forms, and forward-looking. At the agency, an eCRF is now a compilation of lots of data, possibly all of the data about one patient.
On page 3, the guidance states: "The eCRF is a vehicle used to assemble all the data from different electronic- and paper-based systems and makes it possible to capture and organize these diverse data in a manner that satisfies the study protocol and that enables the data to be systematically reviewed and analyzed." Got it? The key word there is "all." Are all of your organization's eCRFs "vehicles"?
On page 11, the FDA adds: "The eCRF is the electronic document containing all data elements on a study subject that the investigator has reviewed prior to release to parties in Tier 3 (e.g., the sponsors, CRO, institutional review board)." The word "all" is used again.
As conceived by the FDA, an eCRF is an uber-record containing both electronic and paper information. This is a legitimate perspective. Regulators could not allow important data to escape their review simply because it was collected in one medium but not another.
According to the FDA, sponsors are not to be in control of eCRF systems. Rather, investigators are to be the custodians. Really? In practice, many investigators are passive recipients of what the sponsors' systems give them. Any romantic notion that an investigator will control the eCRF is akin to saying a passenger can control a jetliner. The FDA may need to get out of the office a bit and talk to investigators about what they truly control.
Most of today's clinical trial systems are already collecting all the data that the FDA wants in an eCRF. But by definition, not all computer systems have necessarily been recording every last paper document, every scrap of information ever jotted down. So if the FDA document stands, some clinical trial systems may have to be tweaked to describe more of the paper flow in clinical trials.
The FDA is also spending considerable time on "identifiers" in its esource guidance. "Identifiers" are short notations that signal changes to data. Identifiers are already being used widely; you can consider them time-stamps. But in the eCRF, some identifiers may need to be applied to paper forms in a more unified, consistent way.
No matter how the language of the draft guidance is amended, it could be prudent to ask your vendor if both paper and electronic data can be aggregated to conform to the FDA definition of eCRF. At the moment, the industry is more accustomed to describing a "casebook" of a patient's CRFs as a data export in its own right. And "identifier" is not yet a routine term. Another question: are the necessary identifiers present for all types of data in your study?
It's also worth noting that the new esource guidance refers to electronic health records (EHR) in several locations. This is auspicious. The EHR is a part of the modern medical environment. At some future date, the EHR will become a routine aspect of clinical trials, too.
But the agency appears to be postponing a tough discussion of the interplay between existing clinical trial systems and EHRs. Because two different worlds of technology vendors supply these systems, it is not at all guaranteed that they will be able to interoperate.
In some cases, companies supplying the same EHR system to two different doctors may not be able to make that system communicate with another database. Getting disparate EHRs to shake hands is another challenge, as is connecting multiple clinical trial systems to multiple EHRs.
All this is not, strictly speaking, the FDA's problem. It's a vendor issue. But the FDA's unwillingness to set down a few rules will hasten the day when clinical trial data are lost because of incompatibilities between clinical trial systems regulated by the FDA—and EHRs regulated elsewhere.
Making matters worse, there is a growing pile of informal database guidelines and interoperability recommendations from four or five nongovernmental organizations. And overlapping sets of privacy rules. So the EHR-eclinical landscape is a mess.
Future EHR Issues
Did the FDA create that mess? No. But the agency could have said more to start the cleanup. It's a pity the agency did not provide more detail about how data from EHRs should be handled inside clinical trial systems. Are electronic data alone ever sufficient? Will a modern pharmaceutical industry be obliged to use paper documents forever, like Amish farmers guiding their horses down rural highways?
In the guidance, on page 6, we can find this about EHRs: ''The sponsor and/or investigator may also be asked during the inspection to provide information on the ability of the software to transfer accurate and complete data from the electronic health record into the eCRF. Algorithms for data extraction should be described in the study protocol or in another document that includes data management details. For example, some patient data in the electronic health record, such as concomitant medications, may change with time. The procedure for selecting the appropriate data element should be described."
OK. But that raises more questions than it answers. What if the algorithm changes during a trial? What if some data flow back from the clinical trial system to the EHR, and then need to be corrected or amended? What if the data in the two systems are discrepant? What if details about the EHR are unknown at the time of the start of the trial, as is likely to be the case in a country with hundreds of different EHRs?
In short, there is a great deal for the FDA to sort out before the pharmaceutical industry can use esource data with confidence, and before clinical researchers can use the EHR systems that they are familiar with from daily usage. d9A2t49mkex
At a time of economic uncertainty, one other observation may be helpful. This is not just an arcane linguistic discussion. The stakes are large. The inability to use EHRs in industry-sponsored research could be another nail in the coffin of U.S. scientific competitiveness. As more American physicians turn away from pharmaceutical research in frustration over the sheer number of systems they must use, countries with a single national EHR will have a significant and perhaps permanent advantage.d9A2t49mkex