XML Takes on a Documentation Nightmare

by Lowell Rapaport

Name any industry and you're likely to encounter complex technical publishing challenges somewhere in the content mix. Financial services firms create investment prospectuses. Manufacturers develop internal product diagrams and assembly instructions as well as customer-facing product literature and instruction manuals. Software companies develop documentation and help manuals. Pharmaceutical companies produce reams of trial and test data in order to win approval, and then they carefully craft literature for doctors and patients once the drug reaches the market.

What all these applications have in common is that they involve highly structured documents that undergo rigorous review and revision - often under the watchful eye of regulators. What they also have in common is a huge opportunity for greater efficiency through the use of XML.

Faced with myriad documents and drawn-out approval processes, pharmaceutical companies have been in the vanguard of XML-based publishing. Nowhere is the need more pronounced than in Europe, where drug companies must generate container labels, inserts, advertisements, product characteristic summaries and other documents in no less than 13 languages. Pharmaceutical companies may have to create and gain approval of more than 600 separate documents for each medication.

"The situation is untenable," says Andrew Marr, director of electronic documentation at GlaxoSmithKline's European headquarters in Middlesex, England. "European pharmaceutical companies currently create documents in Microsoft Word. Whenever there is a change in a drug's documentation, all affected documents have to be manually changed, plus each changed document has to go through a separate re-approval process."

Between the regulatory scrutiny and the sheer volume of documents involved, long lead times and excessive costs are the norm, and there is no guarantee that all the documents will be changed uniformly.

In hopes of speeding approval and cutting costs while also improving accuracy, GlaxoSmithKline has taken the lead in a Product Information Management (PIM) project. The initiative is supported by both the European Federation Pharmaceutical Industry Association (EFPIA), an industry trade association for pharmaceutical companies, and the European Medicine Evaluation Agency (EMEA), the body that tests and regulates drugs for safety and effectiveness.

"In Phase One, we formulated an XML document type definition (DTD) that combined all the different elements of content found in a drug's documentation," says Marr. "This one master document contains all the information about a drug in one place."

The concept is simple. If all the data about a drug is condensed into one XML document, then only one document needs to be changed and approved. All other documents about a drug are just subsets of data from the master document.

In initial trials, 10 different companies put their documentation into the master DTD format, but according to Marr, the project immediately ran into trouble.

"The DTD-formatted master document was unlike any document anyone had worked on before," Marr explains. "Users didn't like the XML authoring tools and were confused by a single large document that included a lot of additional data, much of which was unfamiliar to them."

Users who were used to authoring familiar documents with a particular subset of data were confused by the breadth and depth of detail in the master document. For example, content authors in the marketing department didn't know what to make of the scientific data.

"To make the PIM work in the real world, we had to combine an easier-to-use XML authoring tool with a database," says Andrew Laughland, an IT business consultant with GlaxoSmithKline. "We wanted to let users create the documents they were comfortable with, but as the documentation takes shape, [the data] is transferred into a database where the master document is generated."

GlaxoSmithKline enlisted the aid of Arbortext, the Ann Arbor, MI-based maker of Epic XML authoring software. "The Epic editor offers a user interface similar to Microsoft Word and other word processing programs," says Laughland. "With it, we were able to create standard forms and document views familiar to users."

According to Laughland, the Arbortext editor was easily integrated with Oracle IFS, an Internet-based file system that lets users access and revise content online. XML-tagged content is imported into the Oracle database in which the master document is generated. Once complete, the master document is forwarded to the appropriate government agencies where it can be broken up into smaller documents for approval.

Oracle IFS supports version tracking of individual pieces of content as well as entire documents. When a change is made to an individual piece of XML-tagged data, only the changed content needs to be sent to government agencies for approval. Since all published documents are derived from the master document, the approval process has been streamlined from the chaos of multiple versions of separate documents. And since all documents are generated on the fly from the master document, updates can be published almost instantly.

At this writing the PIM project is still under review. "We are working on how different components, such as XML authoring tools, database tools, content management tools and workflow, are integrated to make a complete system," says Laughland. "When we finish the project by the second quarter 2003, we hope to have a demonstration system that can show the rest of the European pharmaceutical industry how to reduce the time and labor needed to document drugs."