GlaxoSmithKline’s Safety Dashboard

Aug 15, 2005 | Congratulations. Fabuluoxx, your new weight-loss drug, is a blockbuster. But just as clinical trials of Fabuluoxx predicted, the drug causes a benign side effect: hot ears. Every week, hundreds of physicians are duly filling out forms to tell you about healthy, hot-eared patients. What to do? How to be certain a few hot-ear cases will not progress to deafness or worse?

For serious, previously unknown side effects (strokes, deaths, heart attacks), the FDA must be notified within two weeks. But most accounts of diarrhea, rashes, headaches, dry mouth, and the like are sent to the FDA more periodically, in batches. For large companies, the sheer volume of data is a problem. Thousands of reports for one successful drug can make it difficult to focus an organization’s biomedical expertise on the most clinically significant reports. That makes it hard to say if a fatal heart attack in Ohio is a harbinger of another Vioxx mess — or an unfortunate but statistically meaningless event.

Power Sifting
GlaxoSmithKline has found a better way. It started with a massive database of 750,000 reports of side effects that goes back to the 1960s. The database was designed to handle individual-case safety reports and generate paperwork to send to U.S. and other regulatory agencies.

This year, Glaxo bolted an Online Signal Management (OSM) system to the database. It’s a Web-based, automated data-mining and visualization application that uses a state-of-the-art statistical algorithm blessed by the FDA.

Using software developed by Lincoln Technologies, the OSM system won the 2005 Best Practices Award for clinical technology. June Almenoff, vice president of safety evaluation and risk management at Glaxo, notes that with 65,000 reports to process every year, the company’s 130 safety experts needed ways to prioritize. “We have automated a number of alerts,” she says. “You want to have a clear plate so that you can focus on what’s really important.”

One aid: colorful sector maps, originally developed by Lincoln for FDA safety reviewers. These graphics are a bird’s-eye view of the data that shows numbers of safety incidents, organized by severity and organ system. The sector map graph is deeply interactive. “Users can click on any particular square and drill into the case and see what led to that increased signal score,” says Almenoff.

Drug-event pairings (say, constipation and Fabuluoxx) known to be of minimal concern can be manually filtered out of future consideration. But the system is programmed to re-flash an alert if the frequency or severity of the hidden events should change.

The underlying Glaxo adverse events database has always tracked side effects that were (or were not) anticipated in the drug’s label. It also assigned seriousness scores to each incident. That enables the OSM system to generate monthly, individualized reports for every Glaxo safety expert. Even mundane symptoms or complaints may show up if they are new, frequent, or not mentioned in the drug’s label. The reports are generated automatically and show 6-month trends in the data. Says Almenoff: “We’re using that knowledge in a new way. The system says, ‘In the past month, these are the events that are not labeled and are serious.’ We really sort out a lot of stuff. You can focus your work based on those priorities.”

Electronic Documentation
The system also fosters an institutional memory. In the past, towering stacks of paper files had made it difficult for one person to remember how a particular drug-event combination had been scrutinized two years before. Now Glaxo’s OSM system records every choice, every judgment. “It all stays with the product,” Almenoff says. “A manager can look in and monitor all the decision-making that is being done with a product.”

The current system, Almenoff expects, may be especially good at detecting (or ruling out) serious problems with popular drugs that have large volumes of similar reports. Says Almenoff: “When you have 10,000 reports, or 30,000 reports, as we have for some drugs, it’s very helpful. The system integrates the old-fashioned methods of looking at individual cases with new visualization and statistical methods.”

Before, she says, the reviewers’ workflow was heavily dependent on clinical judgment and intuition. Those factors are still present. But the computer can combine and sift the data first. “Everything is done online. The individual case is not what drives signal detection. Now you look at the data more in aggregate.”

SEEING THE SIGNAL: Glaxo connected its global safety database to new software to manage and analyze adverse events, including this interactive digital map of reports organized by organ system.