IBMs Search Engine Sifts Drug Trial Info
On March 22, IFPMA launched an Internet search engine that finds information from a collection of databases and allows searchers to input criteria in five languages.
Both medicines and diseases are known by a variety of names, both across geographical regions and medical specialties, so searches that use only one term can routinely miss relevant information.
The clinical trials portal automatically expands search terms to cover synonyms in medical jargon and other languages and even helps correct misspelled words.
Patients and physicians can also look for trials under way near where they live and learn about the criteria for participating.
IFPMAs clinical trials portal will not host data itself, said Harvey Bale, Director General of the IFPMA. Instead, it will comb through available databases.
"It doesnt replicate a single thing thats being done elsewhere," said Bale. "It brings under one roof a variety of information that was elsewhere only at separate sites."
Currently, the portal links to some 15 sites, including clinicaltrials.gov, centerwatch.com, and clinicalstudyresults.org. Over 88,000 pages have been indexed. The number of sites and indexed pages are both expected to grow.
Automated algorithms and a team of domain experts were used to "teach" the system how to pull up relevant information, said IBMs Mark Andrews, who helps lead the content discovery developing the search engine.
Text imported from medical dictionaries and drug industries was used to help the search systems learn appropriate connections.
The text processing components of the search portal use an open-source framework, UIMA (Unstructured Information Management Architecture).
Its open platform means that a variety of analytic approaches can be used to pull out and apply the appropriate information, allowing "the system to take on a better understanding of the underlying content" of life sciences and drug discovery.
Andrews says that such "verticalization" of search technology can make information much more useful.
"Something as simple as a keyword search would never support the need of users trying to find information about trials," he said.
"Youd either miss information or end up with a lot of irrelevant noise."
Besides just indexing searchable terms, he said, technology must be able to integrate information, extract relevant information, and guide searchers to the relevant information, Andrews said.
Part of the push for the database was a call for drug companies to publish more of their clinical trials results, following reports that GlaxoSmithKline had withheld unfavorable results for its antidepressant.
Bale said that the portal goes "beyond the need to respond to transparency issues" because it will help find not just results of completed trials, but also ongoing trials that doctors or patients might want to participate in.
However, said Beale, the portal would encourage drug companies to publish their results, partly by making it easier to realize that a trials results have not yet been published.
Over the next two months, IFPMA intends to promote its portal through Yahoo and Google with sponsored links tied to specific search terms. An earlier, less sophisticated version of the search engine received about 15,000 visits per month.
Bale said that there are already plans to improve the portal by making the search interface more intuitive and expanding the languages covered.
Check out eWEEK.coms for the latest news, views and analysis of technologys impact on health care.