Skip navigation

CDISC Blog

Clear Messages from FDA CDER and CBER

February 10, 2010 at 7:26 AM by cdisc


There are some very clear messages coming from FDA’s Center for Drug Evaluation and Research (CDER) and Center for Biologics Evaluation and Research (CBER).  As of a meeting with CDER and CBER Representatives and the CDISC Executive Committee on 1 February, here are a few highlights:

  • Use CDISC Standards (SDTM and ADaM) NOW in eSubmissions to CDER.*
  • If you want high quality eSubmissions (which will be reviewed better and more quickly), start with CDISC CDASH for your case report forms when you plan your study.   
  • The expected transport format for the foreseeable future is SASXPT with Define.xml.

*NOTE: CBER plans to accept eSubmissions with CDISC standards sometime in 2010 but please contact Amy Malla about this in advance.

 

The CDISC business case (the full version is in the Members' Only Area), which was completed several years ago, indicates that there is significant time and money saved if CDASH standards are used for study data collection (CRFs).  When CDASH is not employed, many have found that it does not work well to map (‘shoehorn’) data collected in other ways into SDTM at the end AND it is costly and time-consuming.  Yet, despite this knowledge, I keep hearing about all of the legacy data conversion going on; while contractors doing the conversions make money, sponsors complain that implementing the CDISC standards is expensive….Sure it is, if you do mapping at the ‘back-end’.   Instilling quality at the ‘front-end’ of a process has proven effective in several other industries (e.g. ‘lean production’).  In this economic climate, I keep thinking we will see more ‘lean clinical research’, not only to yield higher quality eSubmissons but also since some have told us that they can start a study in 70-90% less time if they use the CDASH standards! 


Okay, back to the topic at hand….  For the past several months, since CDISC did a Stakeholder Analysis late last summer on SHARE, we have been in ongoing discussions with representatives from CDER and CBER.  They have been seeking input about why they are not seeing more eSubmissions using CDISC standards.  One reason is that some were waiting for the HL7 messages.  Well, the CDER and CBER Representatives said to us, we need to walk before we run.  We need to have standard data NOW so that we can see what the issues are in using the CDISC standards (regardless of the transport format) and we can then provide helpful feedback to those who are preparing eSubmissions.  And, we should not surprise the industry, we should be predictable in what we do; hence, we will not be switching to a new transport format in the foreseeable future….and certainly not until it is thoroughly tested and shown to be ‘net better’ than what we are doing today.   (In fact, the PDUFA IT plan will be corrected; it is erroneous to communicate that in 2013 FDA will only accept HL7 transport formats.)


One thing that the FDA reviewers have seen in eSubmissions to date is that there is a need for augmenting the CDISC standards. The safety domains that are available apply across all studies; however, there is always ‘disease-specific’ data left over and users don’t know what to do with it.  Many times it ends up in the Supplemental Qualifiers.  Therefore, they want CDISC to speed up the rate at which we develop the standards around these therapeutic areas.  More information will be coming on this topic in the near future (and it is very closely linked to the SHARE work).  Another comment indicates that providing CRF data with submissions may be helpful (perhaps another opportunity for ODM?). In the interim, they recommend that anyone preparing for an eSubmission meet with their review division well in advance to work out how to handle these ‘extra’ essential data items.


In conclusion, we learned in our meeting with CDER and CBER Representatives that “use of the CDISC standards, refined and developed as needed, will result in high quality submissions that will be reviewed more quickly; as such, the likelihood of an approval on the first review cycle will increase”.


There is a Summary of the recent 1 February 10 CDISC Executive Committee-FDA CDER/CBER meeting in the Members’ Only Area of the CDISC website.   Those of you who have not yet joined CDISC may wish to do so soon; there will be more interesting developments in the near future. This will no doubt prove to be an interesting year!



TagsCDASH, CDER, SDTM, CBER, FDA, ADaM,



Email Post

Comments


 Jozef Aerts February 10, 2010 10:07 AM
My \"crystal ball\" has foreseen this. In my last newsletter (see http://www.xml4pharma.com/Newsletters/) I already predicted that CDER and CBER would have enormous problems implementing HL7-XML, and that it still needs to be proven that that format can port SDTM data anyway.
Part of the problems with SUPPQUAL come from all the limitations of SAS transport, and can be overcome by a much simpler XML format for SDTM submissions than HL7-XML provides.
At least three vendors use an (intermediate) XML format for SDTM based on ... ODM. So why not use that as a basis for SDTM submissions?
The knowledge how this can be done is already there, and such a format combines magnifically with define.xml (also based on ODM).
 Mark Wheeldon February 10, 2010 1:44 PM
I agree wholeheartedly with Jozef\'s comments about three vendors using an XML format for SDTM. This is not standardisation ... each of those vendors knows how to create what should loosely be termed as define for data .. we each all have our own \"proprietary\" implementations however!!

A small collective effort between all in the XML Tech team with CDISC\'s endorsement would have SDTM data delivered in an XML format very rapidly. Right now it cannot be sensible to have metadata that reflects data in a completely different format (XML vs SAS Transport Files).

The time for define for data is now - it would take only a short while to implement (the XML team know what the schema should look like), restrictions on define metadata would be lifted (imposed by SAS Transport File Format), new review tools could easily be implemented on the back of an XML backbone, furthermore the FDA would get knowledge of data in an XML format before HL7 was introduced.

Seems like a win-win for all ....
 Tom Marciniak February 10, 2010 2:54 PM
Let’s inject some reality into this discussion: The experience with SDTM in the CDER review divisions, in particular cardiorenal, has not been favorable. I have done 57 NDA submissions, three of them SDTM. Five of the 57 submissions had severe data problems, including ALL THREE SDTM submissions. ALL THREE SDTM submissions missed their PDUFA goal dates because of the data problems; one of them we even refused to file the first time around. My cardiorenal co-reviewers also experienced severe data problems with a fourth SDTM submission and considered refusing to file. Please consider this track record when planning your NDA submissions. “High quality” and “reviewed more quickly” are goals and potential right now, not reality.

Thomas A. Marciniak, M.D.
Medical Team Leader
Division of Cardiovascular and Renal Products, CDER

 Tom Marciniak February 10, 2010 6:49 PM
The following disclaimer applies to my previous comment: No official support or endorsement of this comment by the Food and Drug Administration is intended or should be inferred.
 Theresa Mullin and ShaAvhree Buckman February 17, 2010 7:08 PM
Dr. Marciniak provides some important observations. First, submitting data in standardized form is not a substitute for clear evidence of safety and effectiveness that NDA data must provide to gain marketing approval. Second, the CDISC SDTM and other CDISC standards need more refinement to meet the needs of FDA reviewers. Realizing the promise of greater process efficiency and predictability requires these refinements. These include but are not limited to specification of the clinical endpoints needed for review in each therapeutic area, further specification of safety endpoints and other clarifications (e.g., to address the data that sponsors currently put in the SUPP QUAL section) to remove the uncertainty and variability in sponsor interpretations of the current CDISC implementation guide. Third, once the data standards are fully specified to meet reviewer needs, their value will critically depend on whether the sponsor submissions actually comply with these data standards.
FDA is beginning to implement a systematic approach to collecting the feedback from reviewers needed to identify required revisions to the current CDISC standards and implementation guide. Such an approach will support the continued improvement and evolution of CDISC standards in an organized way in order to meet reviewer needs.
Additionally, it has been FDA’s experience over the last several years that sponsor implementation of CDISC standards has been inconsistent with even clear rules of the implementation guide. To address this, FDA will be working to model a more effective sponsor-review division interaction that will make FDA’s expectations clear to sponsors.



Theresa Mullin, Ph.D. and ShaAvhree Buckman, M.D. Ph.D.

Co-Chairs, FDA/CDER Computational Science Center




 Arran Oakes February 19, 2010 5:52 AM
From speaking to many of the top 10 pharma companies, there does seem to be two schools of thought - on the one hand, as Becky points out, some large pharmas are trying to map large volumes of existing legacy data into CDASH, when this is inefficient both in terms of time and financially.

On the other hand, some are basically drawing a line under their legacy data and implementing CDISC standards as early as possible by structuring their trial designs with CDASH at front-of-mind.

This seems to be the more efficient and cost-effective approach, with companies striving towards ensuring CDASH compatibility as far upstream as possible, to get the most out of data standardisation.

Of course, it is also important to take Dr Marciniak\'s comments on board, which could be said to boil down to one key message - even if we assume the industry is ready for CDISC, is CDISC really ready for the industry? Some might say there\'s more bugs that need ironing out before its safe to implement 100%.

In the end though, if the FDA openly believe that CDASH standards (ignoring transport formats for the moment) are going to \"result in high quality submissions that will be reviewed more quickly; as such, the likelihood of an approval on the first review cycle will increase\", is it not less of a matter of \"if\" and more of \"when\" the industry will follow?
 Mark Wheeldon February 19, 2010 7:17 AM
Some great comments here. I agree wholeheartedly that the need for specification upfront will lead to the quality and reviewed more quickly goals that we all strive for.

The simplest way to achieve this is to use CDISC metadata upfront to produce a set of both human and machine readable specifications that could be shared with reviewers upfront. For example CDISC ODM (regardless of whether it is holding CDASH content) can be utilised to produce specifications which would aid upfront development of the study (annotated CRFs - both database and SDTM annotations and tabular database specifications.). Furthermore, once these annotations are stored in an ODM library they can be reused from study to study saving time and money in study setup.

The next is to use define.xml upfront. We do this and have use cases where a define is produced and a correponding human specification to clearly articulate what will be submitted in a study. These lead to different uses of define.xml - at study start-up for a specification (again shared with a reviewer to \"iron-out\" differences in interpretation of the standard/clarify choices of SUPPQUALS), during the study to aid legacy conversions and at submission where the original define.xml is refined to reflect the data actually captured (may lead to a culling of unused SUPPQUALS for example) and auto-published as define.pdf and define.html (for ease of human review).

All of these approaches can be introduced today and have been proven to save time and money in traditional trials (where CDASH and CDISC) are not used and even more time where CDISC has been used end to end.

In response to the last point about legacy conversion to CDASH. The CDASH-ODM team have addressed the difficulty of converting from legacy to CDASH very efficiently. No renaming of field names will be required for example - since CDASH item names are implemented as an Alias to the legacy field name. No recoding there just use old legacy names and provide a CDASH alias.

The issues of cost and time may be around structuring of the CRFs and this will be an issue whether CDASH is adopted or not. SDTM is predominantly \"vertical\" in nature and raw operational data is captured \"horizontally\" mainly (since many EDC systems do not support these structures). Yes there may be some restructuring of forms but this restructuring will pay huge dividends in terms of reused mappings at the far end - with CDASH and the ODM + SDTM standards it is a lot easier to create the source CRF, target domain and see how the data flows upfront.

If we only had a standard for the storage of this mapping metadata we would be the whole way there ... please CDISC !!!!

We have shown with a wide variety of data from a variety of clients that end to end CDISC is a reality today. I do not think bugs stand in the way of implementation - it is interpretation of the standards that are a road block. Define is so underutilised and when discussing it with our clients we see a wide array of different assumptions and interpretations. None of which are wrong and all within the relevant implementation guides.

If we get the interpretation of the datasets clearly articulated in define upfront and can share this with a reviewer upfront we will get a lot better quality with the delivered SDTM datasets. This has been proven in the real world between sponsors and CROs and think a similar arrangement would work well between sponsors and reviewers?


 Arran Oakes February 19, 2010 12:17 PM
Mark - its interesting to hear how your CDASH-ODM team have handled the approach to legacy mapping - seems almost too easy! (too good to be true?)

I\'ve spoken to quite a few solution providors and yet not heard that approach. Very innovative! If all the big pharma companies could map their legacy data over quickly and easily, could it not push the FDA slightly further in the million-dollar question: will they enforce CDASH?

I think industry-wide standardisation could be catalysed from either side, whether it be the FDA enforcing it or the industry coming together and agreeing on it.
 Jozef Aerts February 20, 2010 3:41 PM
The approach for annotating CRFs with CDASH metadata (using the Alias element) has indeed been developed by the CDASH-ODM team and has successfully been tested in some IHE profiles. For example, it proved to be the key enabler to automatically populate eCRFs according to CDASH (or containing some CDASH subforms) with information from electronic health records (EHRs). It worked equally well for EHRs from HL7 CCD, as for EHRs from OpenEHR.
The CDASH-ODM forms using this \'Alias\' mechanism will probably be published in May.
In the mean while you can already have a look at a short \'wiki\' article I have written and which can be found at:
http://www.xml4pharmaserver.com/XML4PharmaWiki/doku.php
and look for \'Annotating ODM with SDTM and CDASH information\'. This can already be used by early adopters especially for annotating legacy CRFs, and is especially interesting when archiving these studies using ODM.

 Kurt Hellstern April 20, 2010 10:39 AM
The technical discussion above is interesting. However, for a small CRO, the discussion about the quality of submissions is more important. Therefore I would like to come back to this part of the discussion. A detailed feedback from FDA reviewers would be very helpful for the data manager’s and programmer’s future work. We all make our own interpretations about what we read in CDISC documents and hear at our Interchanges – but we do not really know, how we may fail in the perception of a reviewer at the other end of the communication. Reality is still the best teacher and errors made by others can be very helpful in identifying pitfalls in our ODM to SDTM or legacy data to SDTM mappings. Thus, I would like to encourage the FDA reviewers to show us examples of what can go wrong and the CDISC organisation to publish this valuable information on the web. Thanks in advance!
 temporary health insurance December 26, 2011 1:37 PM
The collaboration that is forming between research teams and healthcare industries is so exciting to me. I am also happy to see such a proactive approach to it as well. It seems to becoming more well received on both ends as well. I look forward to seeing further development on this.