CDISC an industry consortium dedicated to establishing clinical trial data standards



CDISC (Clinical Data Interchange Standards Consortium), a volunteer, multidisciplinary, non-profit organization backed by 38 industry sponsors, was established to develop such standards to support the electronic acquisition, exchange, submission and archiving of clinical trials data and metadata for medical and biopharmaceutical product development. The group's mission is to lead the development of global, vendor-neutral, platform independent standards to improve data quality and accelerate product development in the pharmaceutical industry.

According to Rebecca Kush, CDISC founder and President, the group started with no backing or financial support from the industry when two friends, Bron Kisler and Tom Tunstall, who were coming out of other industries, asked about pharmaceutical industry standards. They were entering this area from the technology end and they said: Where are your standards?'s I started laughing, Kush said. It was the middle of 's97 and my honest answer was that you might call SAS a standard, because everybody eventually puts things into SAS for the FDA, but that's it.

The trio, who agreed to try to do something about the lack of standards, began with a conference workshop for DIA called Dynamic Data. We had a panel discussion and people just about threw tomatoes at us, because they said: Are you kidding? We can'st even get standards in our own companies's, Kush said.

But the team managed to sign up 25 people for a meeting held in Washington, D.C., so that FDA representatives could attend. We defined a vision and objectives, Kush said. We got a couple of teams set up and decided on a name. One person leaving the meeting said it was a great meeting, but that the effort would suffer from viability. I was determined it wasn'st going to just die.

We floundered for almost a year, because no one could figure out how to wrap their arms around this big problem, she said. We had two groups a glossary or nomenclature group and a group looking at data models.

The glossary group actually made fast headway by borrowing terms from other industry glossaries and a process from Xerox. But the data models group, according to Kush, was struggling. What finally became obvious was that we had two sub-groups who weren'st talking, she said. We had a group who wanted to look at data the FDA was going to receive submission data and we had a bunch of technology providers who wanted pharma companies to give them some kind of data standards to work with so they could better serve the industry. And they weren'st the same approaches. One group was talking XML and IT and the other group was saying: What are you talking about?'s

Kush began working with the technology vendor group while Dave Christiansen and Wayne Kubick organized the efforts of the submission data group. The FDA had just released its guidance for electronic submissions, so the submission data group began by developing metadata models that could augment that document. The group developed what they called a straw man metadata model that stayed away from the proprietary data standards other groups had tried to use.

They didn'st get too close to what everybody felt was proprietary, Kush said. They backed up a step and worked on the metadata. They said: We don'st want to tell everyone how to do this, but we want to give you a model to follow. We don'st want to step on any toes and we don'st want to compromise the scientific integrity of the data that you are exchanging. Our data isn'st like that for an ATM machine we'sve got research going on and every protocol is different.'s

According to Kush, that's why other standards groups had failed. They wanted to start at the variable name and nobody could agree, she said. So with metadata, you back up a step and say I don'st care what you use for the variable name, but attach this information about it so that when the FDA gets it they know what they received.'s

Dave has a great example of the need for metadata, Kush continued. It's essentially the reason that the Mars orbiter crashed, because one group of scientists didn'st know they were working with English units and the other was working with American units. That's the metadata.

A group of 20 people from CDISC, most of them from pharma companies, developed metadata models for all of the safety domains that go into electronic submissions and presented those to the FDA in October last year. The FDA has considered our input, Kush said. They'sve gone through it with the Submissions Data Standards team and evaluated it and will be considering it as they write new guidance documents and update existing ones.

The vendor data group also has made good progress. In 1999, two independent groups developed XML models for handling clinical trial data and made those models freely available to the industry. The groups were Lincoln Technologies in conjunction with PHT and, in a separate effort, Phase Forward.

They opened their models up and we held a meeting in September and formed a group of seven multidisciplinary's individuals to go away and technically analyze these models and come back with a recommendation for standards to support data acquisition, Kush said. They came back and recommended a consolidated CDISC model that would use parts of both of those and we came out with a version 0.8 of an XML DTD that was posted on the CDISC Web site.

Then we pulled in a group of about 10 reviewers and asked them to give comments on it and we presented that at an industry data management meeting that November. Last September, we published Version 1.0, which is an expanded model. They have added some pieces that deal with audit trail and archive issues per the relevant guidance documents, such as the guidance for computerized systems used in clinical trials.

A new Operational Data Modeling team of 10, including some people from the original data acquisition team but also some new members, is scheduled to have its first face-to-face meeting this month to address a list of goals to augment the model and begin testing it to demonstrate its usefulness. Wayne Kubick is now leading this team.

In addition, CDISC has established a Laboratory Data team to make recommendations on how to expand the operational models to incorporate clinical trial data from central labs, a high-priority need identified by CDISC participants in a group survey. New teams have also been established to examine analysis data sets with the FDA, to support testing of the models and to educate people about the models and how to use them.

Until February 2000, the group operated with no official funding as a completely volunteer effort. For two and a half years we did everything on a volunteer basis, Kush said. We didn'st have a penny in the bank. Our companies were sponsoring most of the travel, so we tried to combine our meetings with conferences people were attending. We couldn'st incorporate because we didn'st have money to pay lawyers, so we had to choose whether to merge with another standards group or form our own entity.

In November 1999, at an industry meeting on electronic data capture in Durham, North Carolina, we held a CDISC meeting with several round table discussion, including one on the future of CDISC, Kush said. At the end of this meeting, the whole table of about 12 people said: Becky, you might call this a coup, but we'sre setting up an organization and PHT is going to pay the legal fees and IntraLinks is going to put up the first Web site .'s And it was done.

We received a lot of support from Quintiles and certain pharmaceutical companies, she continued. They pulled together representatives from10 pharmas, and the CDISC Steering Committee presented a business plan on February 17, 2000 (the day after the non-profit organization of CDISC was incorporated). We basically said: If you
want to see this effort grow and go somewhere we need some funding.'s

By October, when the newly incorporated group had its first Industry Advisory Board meeting, CDISC had signed up 33 corporate sponsors. It now has 38 sponsors.

The challenge now is taking all of this interest, good will and funding and making sure we move the standards forward, Kush said. We don'st want to let the fact that we'sre trying to put an organization together slow that down. The refreshing part of these standards is that all of these companies have realized that it is not wonderful to have proprietary data models for each organization, especially with the recent trend towards mergers.

Kush believes the organization's biggest hurdles are getting people to begin using the models and to participate in helping refine them. Everybody can look at the models and provide feedback through a discussion facility on the Web site. And the submission model can be used in fact, there's been a submission done using that model. What we ask is that no one change the models and call them theirs. There must be CDISC ownership's of the standard models, but we want openness so people can use them and comment on them.

To learn more about CDISC, visit the organization's Web site at cdisc.org.