We're quick to pop a pill these days. But we rarely think about the pain involved in getting that pill to market. The painstaking research and development that goes into adding a new drug to the shelves of your local pharmacy often requires a collaborative effort far beyond our imaginations.
Drug companies must conduct broad, expensive trials before their products ever get to the Food and Drug Administration (FDA) for approval. Out of the tens of thousands of compounds that are discovered each year, only 7 percent of them make it into your medicine cabinet. After an extensive development phase, pharmaceutical companies must back up their claims for a drug's efficacy for the safety of the public by submitting mountains of paper to the FDA -- the federal agency set up to evaluate pharmaceutical products before going to market. The FDA's approval process is also a long and detailed one. In many cases, the documentation that companies must provide to the FDA can grow to a ponderous number of pages -- typically 1 million.
In addition to the elaborate development and approval process, drug companies are also facing added pressures from a Congressionally-mandated restructuring of the FDA's review process. The mandate requires that the FDA shorten its review process to just 12 months from its typical 18 to 24 months in order to get drugs to market faster without compromising safety.
So in an effort to move documentation more swiftly through the FDA's approval process, Pfizer has switched to an electronic-based drug submission process. The giant pharmaceutical company's Esub -- Electronic Submission Navigator -- offers a vast departure from its old paper-centric way of submitting documentation to the FDA.
Pfizer's award-winning system has delivered impressive results. It cost $3.2 million to develop and has so far netted the company at least $142 million in revenues due to shorter cycle times. But it has also given the company much more than bottom- line benefits: Esub has changed the way the research clinicians and the IT staff work together.
"Esub has had enormous impact in that it has transformed the way we do things internally," says Dr. George Milne, president of research and development for Pfizer's Central Research Division in Groton, Conn. "Our ability to execute new drug filings has been brought to an unprecedented scale," he says. "It's much more than just an interesting computer system. The tools that Esub gives us will stimulate insight. We expect it to lead to a cascading effect of innovation," he adds.
"Pfizer's results directly affected its industry and served as a role model and a benchmark," says Gregor Bailar, vice president and CIO at the National Association of Securities Dealers and an Enterprise Value Award judge.
Mired in Paper
In mid-1995, Pfizer's researchers were working on the development of Trovan, a new antibiotic drug that was being readied for FDA approval. The potential new product would be the largest anti-infective submission ever received by the FDA.
Typically, researchers and support staff would have produced separate sections of a paper document to report test results of drug trials. Each section would eventually be assembled into a master document, called an NDA -- a new drug application -- that could amount to as many as a million pages, or the equivalent of a tractor trailer full of paper. Once compiled, the NDA was edited, copied and sent to the FDA to start the approval process.
The FDA would take the report apart, send portiions of it to appropriate reviewers, who, in turn, would write their own analysis. Needless to say, document management, revision control and cross-reference accuracy were a challenge for the agency. Working with 20,000-page sections, individual reviewers had to slog through a stack of papers often 5 or 6 feet deep. They would leave questions on sticky notes and send runners back and forth from a large warehouse where the master document was stored. It might take several days or even weeks for a question about a new drug's potential uses or side effects to reach the appropriate source. "If the reviewer found something in volume 99 on page 20,592 that referenced something on page 9,000, for instance, they'd have to send a request to the FDA library and wait for a copy," says Walter Hauck, then an associate director in charge of clinical application development and now director of clinical systems. "The reviewers needed more than the latest version of a specific section of the document," he says. "They needed the latest version of every section of the document."
To solve the problem, Pfizer's staff used Computer Aided NDAs (CANDAs) to build sections of the document electronically. But they didn't provide all the data, and performance was slow. "It was simply the best we could do with the technology at the time," says Hauck.
In 1995, everyone at Pfizer was talking about trying to find a way to build an electronic version of the NDA right from the start. "Everyone wanted nirvana, but no one knew how to get it," says Hauck, who had only recently come to Pfizer. "Back then, there were no good search engines, and the technology just wouldn't support an electronic solution," he says.
But the web was beginning to take off, and Hauck suggested that it might be time to experiment with it. "I was the new guy, so I didn't know that we couldn't do it."
Hauck and his IT team showed a crude prototype in April 1996 to one of the Trovan clinicians, who in turn showed it to Scott Hopkins, executive director of anti- infectives. Hopkins instantly saw the benefit of the project and gave it the go-ahead.
Pfizer's IT team created about one Esub prototype per week, rewriting the code nearly 40 times until it got easier to manage. Yet, not surprisingly, Esub's development suffered some fits and starts. "Everyone grossly underestimated the size of the NDA for Trovan," says Hauck. "We guessed that it would be 200,500, maybe even 1,000 documents -- somewhere in the range of 50MB," he recalls. But the Trovan submission grew to almost 50,000 documents, "close to 180GB," he says. "It was the biggest thing I'd ever seen!"
Yet as the project grew, enthusiasm for it also grew. "The IT team would work for three hard days building a prototype. Then we'd deliver it to the clinical team and say, "What do you think?'" Hauck says. "They'd use it, give us feedback on what they liked, what didn't work at all. And we'd go back and build a new prototype."
At one point, the clinical team members became concerned that electronic files would not faithfully reproduce scientific symbols. So they hired 30 summer interns to painstakingly check through 20,000 documents to ensure that those files were accurate. If that weren't enough, about a month before the Trovan drug submission was due and Esub was expected to debut, the team discovered that one of the regulatory reviewers would be conducting his review out of his Rhode Island home via his own PC. So the IT team had to completely configure another server, including building the content, the feeder system and the business rules, at a cost of $200,000.
After working together for so long, the team members became close. "We had worked our hearts out, and sometimes I wanted to cry," Hauck says. "We ate up several people and caused a lot of divorces with our 100-hour workweeks," he says. "But we had a talented IT staff and a committed clinical team who began to believe this technology was magic."
On Dec. 28, 1996, nine months after its conception, Esub was delivered to its two proud parents. Pfizer's clinical and IT teams shared ownership of the risks as well as the rewards. "Without the IT and business collaboration, Esub would have gone nowhere," says Hauck. "We all worked together to put an incredible amount of information on the web, and we made it searchable in under a second," he says. "And all it takes is the push of a button."
In hindsight, everyone at Pfizer was happy that Trovan was the project's firstborn. "It was so big," says Dr. James Milson, vice president of information management at Pfizer. "We got to solve the tough problem upfront. Then, when we rolled out Viagra in March 1997" -- a much smaller project because Viagra had only one indicated use -- "we didn't need to worry about scale."
Since its approval Dec. 27, 1996, Trovan has run into several roadblocks. In early 1999, several patients who had taken the drug showed signs of liver problems, and consequently the FDA placed tight restrictions on its use. Esub, however, has not suffered the same fate, and its benefits go way beyond a typical return on investment.
Esub has become a companywide data-sharing system that is also being considered by the FDA as a benchmark for other drug companies' submissions. "It doesn't hurt our competitive advantage to give the tool away," says Hauck. "It's only software, you know, a couple of lines of Perl. It's the content that's important," he adds. "Our competitive advantage is that we know how to build quality dossiers in real-time."
Esub has also unwittingly created a heightened role for IT within the Pfizer organization. "We now sit right at the business table and no longer just react," says Hauck. By working collaboratively with its business partners, Pfizer's IT team constructed a system that provides a global view of the status of a trial or application process enhances Pfizer's competitive advantage by linking drug researchers around the world; its Esub system has attracted business partners, including other drug manufacturers that seek to forge strategic alliances with Pfizer to help market and distribute their drugs enables Pfizer to penetrate world markets much more quickly by filing concurrent submissions in different countries gives the company the ability to deliver five new drugs every 12 months, an industry-leading rate features an electronic table of contents to negotiate the forms of an NDA, with full title and text search capabilities allows portable review with a full-featured system -- a boon to the FDA, which frequently uses outside consultants to speed the regulatory process.
Hauck says that getting the business users -- many of whom are medical doctors -- to sign on with the Esub effort was easy. "Doctors are competitive," Hauck says. "So we'd say, "So-and-so on the Trovan trial team is really benefiting from Esub,' and the next thing you know, another business unit would want to be involved." The number of users of the system has grown to 2,000 worldwide at both Pfizer and at regulatory agencies. In addition, the Esub repository has grown to 5 terabytes, with roughly 1 terabyte of new data added each quarter.
Esub has elevated IT's role within Pfizer and helped the department align more closely with the business side. Now the IT department is regarded as a close partner in building drugs for the market. "We don't have a centralized culture," says Milson. "But when it comes to Esub, we are working together. We get the data we need far more quickly and easily, and we're building partnerships that are important for innovation."
The real value of this system comes in some intangible ways, says Hauck. "This was not a static project," he says. "It helps us, and it helps the regulators make sure that the work we do is as safe and effective as possible. It makes us very proud. And it greatly enhances the role of IT in all we do," he says. "It really has no end. It just has endless possibilities."
Anatomy of THE SYSTEM
HARDWARE Small servers, such as dual-processor Sun Microsystems workstations, maintain content. Client workstations need only an HTTP browser and a PDF reader to view submissions. Data is analyzed using applications such as SAS, either on the desktop or via an X-windows server.
SOFTWARE Content is stored in the file system using Revision Control Software. The Electronic Submission Navigator software is written in Perl. The Esub software is supported by several Solaris administration utilities and tools and is written in C using the Adobe PDF software development toolkit.
NETWORK Esub's web-based interface allows for mobile review via laptops. Content is indexed for searching using a wide area information search engine. With a combination of caching and cookies, the table of contents can be redrawn within 0.5 seconds, even over low-speed connections
This story, "Prescription Strength" was originally published by CIO.