Sponsors Urged to Use CDISC Clinical Lab Model for Data

Using the “lab model” standard developed by the Clinical Data Interchange Standards Consortium (CDISC) can help clinical trial sponsors transmit large laboratory data sets more efficiently, according to Bob Lyons of Nextrials, who was speaking at a recent CDISC conference.

“It has built-in data structures that are straightforwardly linked to clinical concepts,” Lyons said. “It also has a very flexible transactional model to support data that’s coming in.” The model is based on extensible markup language (XML), a next-generation internet language.

Nextrials, which provides internet-based software for the clinical research industry, has integrated support for the CDISC lab model in its Prism clinical trial management software.

The company has its centralized lab clients send data to it in the CDISC lab model format, and then places it on its own file transfer protocol (FTP) server for evaluation. “Then we unarchive it and chunk it into pieces at the subject level,” said Lyons. “It is also sent to our application server … we can interleave [the incoming data] with other requirements, rather than doing it all at once.”

Growing Number of Centralized Labs

Nextrials adopted the CDISC lab data model because a growing number of clinical trials use centralized labs, Lyons said, adding that “manual entry of lab data is highly error-prone, and if it’s late coming into the database, it’s hard to correct. And customized data loading is not cost-effective.”

Implementing the standard took Nextrials about one-and-a-half months, Lyons said, and now “about 35 percent of all projects, and 90 percent of new projects, are using this facility.” The company processes approximately 358 results per project every day, Lyons said.

In Nextrials’ experience, adopting the CDISC lab model makes data transfer much faster. The average time from specimen draw to entry in the electronic case report form (eCRF) to entry into Prism is now two days, “much less than with other site-entered eCRFs,” said Lyons.

The standard also helps catch mistakes, leading to cleaner data. “The initial error rate is 4 to 7 percent presubmission, mostly subject ID errors and visual test mapping. Submission errors asymptote to zero as the trial continues. Query rates to Prism are virtually zero,” Lyons said.

Resolving discrepancies is not an automatic process, however. “We get electronic notification and contact the sites,” Lyons explained. “If it is the lab’s mistake, they correct it and resend the data. If it is Prism’s mistake, we can query and correct it.”

Mapping Challenges and Visits

The problems with the CDISC lab model in practice so far relate to mapping challenges and visits, Lyons said.

“Subject ID mapping is problematic [due to] manual kit labeling by sites. We are working on having Prism support kit labels with bar coding.”

Visits represent the largest expenditure of time in the configuration process, since “multiple unscheduled visits are hard to distinguish from each other and the tests,” said Lyons.

Also, using the CDISC lab model “sometimes adds a nontrivial amount of time … to the study startup.”

It’s possible to reach the break-even point after doing just two projects by using the CDISC lab model, Lyons said. “Everyone has different numbers, and there are more complex situations, but it’s not hard to implement; it’s pretty straightforward.” — Martin Gidron