CDISC Blog

The FDA, HL7 and a Disappearing Timeline

November 19, 2009 at 8:38 AM by cdisc

It might not have been a presidential debate but the closing session at the North American Interchange was no less intriguing. The session was an FDA panel that opened with a speech from the FDA Deputy Commissioner Dr Sharfstein. The words from Dr Sharfstein were complementary to CDISC and, happily, contained no surprises striking a balance between the new and the present but not delving into too much technical detail.

Then the session changed somewhat. As we would say over in the UK, a frank exchange of views took place.

Sitting on the podium with Dr Sharfstein were Chuck Cooper, Sue Bell and Steve Wilson from CDER, Amy Malla from CBER and Jay Levine and Lilliam Rossario from the Office of the Commissioner.  For those not so familiar with standards development, CDER and CBER can be thought of as users of the standards while the Office of the Commissioner works with CDISC, HL7 and other Standrads Development Organisations (SDOs) to develop new standards.

There seem to be differences of opinion between CDER, CBER and OC about the HL7 message development.  These differences may not be so much about whether we – that is an industry ‘we’ – should move in that direction, but when and what to do between now and such a date.

Now keen observers of this topic will be familiar with the date of 2013. The end of 2012 marks the end of the current PDUFA cycle and that date has also been tied to the introduction of the HL7 messages. Well it became apparent from the debate, and from the JANUS session that took place just before the panel, that the plan for the introduction of the messages in 2013 has been dropped. Also for you close observers, the plan for using an HL7 Version 3 message for the SDTM-equivalent data has been dropped and instead a design based on HL7’s Clinical Document Architecture (CDA) will be used. While messages and CDAs are related this is a significant move. It reflects the emerging US standards for Electronic Health Records which are also moving in the direction of CDA.

There was a debate around why the agency wants to use the HL7 paradigm. Jay Levine mentioned EHRs and the US programme to introduce EHRs. As an aside physicians in the US will be encouraged to adopt EHRs by the use of reimbursement incentives (money to you and me). However, they will only get the money if they are successful in becoming “meaningful users” of certified EHR technology. Now the definition of “meaningful use” is becoming a minor industry, in the US, all of its own but the certification of the EHRs will include the use of standards. These will be CDA based and hence why the move to CDA, by the FDA, makes sense.

Given the developments within the EHR field, there is merit in the decision by the FDA to use the same standards as those used in healthcare. This is true in the post-marketing space, for example the reporting of post-marketing adverse events. And the logic goes, if we use them post-marketing, then why not use them pre-marketing, employing a common information model to ensure the pieces fit. The FDA is already using HL7 mechanisms for the Structured Product Label (SPL) data while the Individual Case Safety Report (ICSR) is well along its development path. Hence the long-term vision of HL7-based exchanges between industry and the FDA. This logic is not new and has been heard before but was articulated during the session by Amy Malla.

So what is the issue? Well industry is confused. Most submissions today, if they are SDTM compliant, are late stage conversions, i.e. the data was not collected in an SDTM compliant fashion but mapped prior to submission. But, however the data got into an SDTM form, the reviewers like it. This sentiment was expressed several times by the CDER representatives on the panel, “We are trying to create a modern review process now. There are challenges to get to the ultimate goal. We don’t want to wait that long; we want to take advantage of the standards now.”

The catch 22 is that reviewers like and want SDTM but we have this HL7 change on the horizon. If I have not “moved” to SDTM as a company, should I wait for this HL7 thing? Am I wasting my company’s investment if I go this way? 2013 was quite close. Now with the date having vanished there is more confusion. As one lady commented from the floor during the Q&A – to the sound of loud applause – “[we are] trying to provide to you what you want but you [FDA] need to make a decision”.

I would like to say that CDISC has all the answers. We don’t. We have been working with the agency to try and understand the impact of the move to a HL7 solution. The data collected will be the same, there is no change there, it is how we are packaging the data for submission to the FDA that is potentially changing. As it was pointed out by another comment from the floor, if we were to introduce the HL7 mechanism in 2013, a lot of the data that would be submitted at that point will have already been collected prior to today. Until we start testing the technical implementation we won’t have all the answers. I have to say that, call it gut feeling, this should work given that it is the same data that will be submitted but that is not good enough as a basis for proceeding. We need hard evidence that it does work, clearly demonstrated with clear guidance on how a submission is to be made using the new mechanisms.

From what was said by those asking questions during the session and thinking about all of this on the flight home from the Interchange, it strikes me that industry needs something it can plan against based upon clear statements from the FDA as a whole. May be a simple framework such as the following four points might work:

• An open and collaborative approach to the development of the HL7 mechanism between CDISC and the FDA (OC, CDER, CBER etc).
• CDISC and the FDA need to test those mechanisms and clearly demonstrate that SDTM-based data that can be submitted in a SAS XPT file(s) form can also be submitted using the HL7 mechanism either using updates to current tools or via new tooling.
• If we can successfully demonstrate point 2 then the FDA should accept XPT and HL7-based submissions in parallel for a period of time (say 2 years) to ensure there are no issues and then move to the HL7 mechanisms.
• If we cannot demonstrate point 2 then we work on the problems until we can. The FDA will commit to continue to accept SDTM/XPT based submissions until a successful demonstration is made.

The second and fourth points are key. Point four is also vital as such a simple statement from the FDA would provide a firm base upon which industry can adopt SDTM today, rather than waiting, without fear that the world will change.

And of course this is all very SDTM focused. There is also ADaM where a similar set of challenges await.

All in all it was a good session and it may just be that we look back at events around this Interchange and this panel session and consider this to be one of those moments where an issue tipped from being difficult to one where a clear way forward emerged.

Your comments on this blog are most welcome.

Dave Iberson-Hurst

Comments

Parag	November 24, 2009 12:47 AM This is a very thoughtful and very well written blog. Although adoption of HL7 would not have sudden impact on the submission method, from standards development perspective, it is essential for industry to get necessary lead time. In such scenario, the framework with four points suggested by Dave are important to understand.
Jozef Aerts	November 24, 2009 4:58 AM Dear Dave, Although I am glad that the FDA abandoned the (i.m.o. bad) idea for HL7-v3 messages as a transport container for SDTM, the 90 degree turn also raises some new concerns. As CDA is based on HL7-v3-XML, my technical (XML) objections remain, but you do already know what these are, so I do not want to discuss these here. As CDA is a very complex format, difficult and expensive to implement, and given the state of IT expertise at CBER and CDER (and of XML e
Jozef Aerts	November 28, 2009 7:53 AM As the system seems to have cut my contribution, you can find the complete text at: www.xml4pharma.com/HL7-XML/Response_to_Daves_blog_2009_11_24.pdf Best regards, Jozef
Jason Rock	December 7, 2009 9:10 AM The difference between a document and message, in my opinion, is more of an administrative issue. For example, my company, GlobalSubmit, developed a proof of concept to load the HL7 study message and SDTM SAS datasets in to Janus 2. We originally developed the proof of concept from the message. With less than 5 search and replace commands, no programming or debugging and a half hour of testing to confirm there were no issues, the proof of concept worked on the new document. The following is from the PDUFA IT 5 year plan from May 2008 3rd Quarter, 2008 •HL7 DSTU Ballot 2008 -2009 •Testing 3d Quarter, 2009 •HL7 Normative Ballot 2009-2012 •FDA accepts both CDISC-HL7 XML •and SAS transport files 2013 and Beyond •FDA accepts only CDISC-HL7 XML The DSTU ballot was delayed for a year.
Jay Levine	December 14, 2009 8:08 AM Just to clarify… FDA encourages sponsors to submit data in SDTM format. Details are provided in the current eCTD Study Data Specifications. There has not been a change in the plan to introduce the HL7v3 Study Data Standard in 2013. The content will be contained in three study data standards (Study Design, Study Participation, and Subject Data). These standards, designed in close cooperation with CDISC, have passed draft standard for trial use (DSTU) ballot, and are currently being tested. Initial tests have been successful. CDISC’s continued support in this effort will ensure a smooth transition to HL7v3 standards. Also worth noting is that Jozef Aerts characterization of the Study Data Standard as a “as a transport container for SDTM” reiterates a common misconception regarding the Study Data Standard. While the Study Data Standard will be capable of representing all SDTM data, it will not be limited by the current two-dimensional SDTM approach to data exchange, and will be able to accommodate a wider variety of study data, as well as more complex study data.
Jozef Aerts	December 17, 2009 4:07 AM There is no misconception from my side. I realize this very well. My objections are against the use of HL7-XML (which abuses the XML standard itself). HL7-XML has been proven to be extremely complex, and extremely difficult and expensive to implement. What the FDA wants to achieve can also be accomplished with much more simple XML structures. Given the XML knowledge at CBER and CDER and at the FDA in general (almost completely absent) I fear that these departments will have huge difficulties with the implementation of HL7-XML. A good indication for this is that 5 years after the introduction of define.xml, CBER has still major problems with doing something simple as validating define.xml files.

Name *

Website

Comment *