Adam Jacobs FICR CSci
Keywords
CDISC, Data acquisition, Data interchange, Electronic data capture (EDC), Electronic health record (EHR), Harmonisation, Standard
CDISC (the Clinical Data Interchange Standards Consortium, www.cdisc.org) is a relatively new player in the world of clinical research, having been founded in 1997, but it is rapidly gaining in importance, and is here to stay. To keep up with some of the latest developments in the CDISC world, I went to the European CDISC Interchange, held in Budapest in April 2009.
CDISC background
CDISC is not a single standard. Rather, the CDISC organisation publishes various standards for various aspects of data handling in clinical research. The most mature and widely used CDISC model is the Study Data Tabulation Model (SDTM), which specifies the structure of datasets used to store raw data from clinical studies, and is the recommended format for submitting raw datasets to the FDA. Other standards include the Clinical Data Acquisition Standards Harmonisation (CDASH), which specifies how data should be collected on case report forms (CRFs, either paper or electronic), and the Analysis Dataset Model (ADaM), which specifies the structure of datasets used for analysis of a clinical study (related to, but not the same as the SDTM).
I was particularly pleased to be going to Budapest, which is one of my favourite cities, and this was my first visit since the EMWA conference there in 2004. Sadly, as is too often the way with business travel, I didn’t have much of a chance to see the city this time, but I did get to return to a restaurant I’d been to before and thoroughly enjoyed, called Nosztalgia. It was still excellent, but I had a strange feeling that it had been better the previous time I went.
Budapest is blessed with some excellent hotels, and the CDISC meeting was held at the Corinthia Grand Hotel Royal, which is definitely in that category. The meeting had two functions: training workshops on various aspects of CDISC were offered on the Monday, Tuesday, and Friday, while Wednesday and Thursday were dedicated to a conference with a variety of lectures on how people were using CDISC in practice.
I started the conference on the Tuesday, and attended training sessions on the BRIDG model (a domain analysis model for clinical research) and the newly-published Protocol Representation Model (PRM), which is based on the BRIDG model. The PRM is still very much in its infancy, having only so far been published in draft form for comments. However, it has great potential to revolutionise not only the way we write protocols but also the extent to which information from the protocol can be automatically passed to electronic data capture (EDC) systems and clinical databases (see comments on this in the ICR/EMWA conference report elsewhere in this issue, Ed.). The next few years should see some exciting developments in this area.
Integrating electronic health records
Many of the lectures on the Wednesday morning were on the theme of integrating clinical research with the electronic health records (EHRs) that are used in clinical practice. Many EHRs conform to the Health Level 7 (HL7) standard, which can be mapped, via the BRIDG model, to CDISC standards for clinical data. The technical challenges of integrating data from clinical practice into clinical research have therefore either already been overcome, or could soon be overcome with only a little extra effort. This gives rise to possibilities such as automatic pre-population of fields in electronic CRFs, by linking to the patient’s EHR, or even querying EHR databases against the inclusion and exclusion criteria for a study, thus returning a list of patients eligible for a new clinical study at a given site and potentially saving considerable time identifying eligible patients. However, despite the great progress that has been made in the technical aspects of EHR integration in clinical research, some significant privacy and legal obstacles remain, which I suspect will be much harder to overcome.
The Wednesday afternoon included a series of talks on the CDASH standard. We learnt how use of the standard had led to considerable time savings in the process of producing CRFs at companies who had embraced the standard. The benefits should be most significant for CROs, particularly if all their clients can be persuaded to use CDASH, as this should reduce the problem faced by many CROs of having to do things in completely different ways for different clients.
Data validation
On the Thursday, I attended some lectures giving tips for how to ensure data are validated against the standards, and how the use of data standards can result in considerable savings of time in statistical programming. One particularly fascinating talk explained how simply preparing a spreadsheet to define which variables should be included in analysis datasets was all that was needed to create ADaM-compliant analysis datasets ready for statistical analysis, once some suitable programs for constructing the datasets from the spreadsheet (which only needed to be written once) had been written.
Sadly, I was unable to stay until Friday, when I would have very much liked to attend the training workshop on ADaM datasets. Luckily, my colleague Flavia was able to go to that so I hope she’s taken copious notes.
Summary
I felt I had learnt a great deal about CDISC standards in my 3 days at the conference. I had been given some great ideas for how various processes at my company can be made considerably more efficient: writing protocols, designing CRFs, setting up clinical databases, and statistical analysis. I hope I’ll be able to go to next year’s conference and give a talk to other delegates about how I’ve been able to implement all those ideas.
