Data Mining Prescribed To Ensure Drug SafetyData Mining Prescribed To Ensure Drug Safety

When a new drug hits the market, it can take years before enough evidence is gathered that the medication poses dangerous -- or even deadly -- risks to certain patients.

Such was the case with Vioxx, which doctors prescribed for several years before data finally proved the arthritis drug significantly raised risks of heart attacks and strokes, forcing drug-maker Merck to yank Vioxx from the market in 2004.

Since then, health care players -- including insurance companies and providers -- are ramping up development of new electronic surveillance systems that can help the Food and Drug Administration and drug-makers identify medication safety problems much earlier by analyzing patient clinical and claims data.

This week, WellPoint -- one the nation's largest health insurers -- revealed it's investing millions of dollars in a three-year project to build such a drug surveillance system in collaboration with the FDA and several academic institutions, including Harvard University, University of Pennsylvania, and the University of North Carolina. The Safety Sentinel System will mine and analyze aggregate claims, lab, and pharmaceutical data from WellPoint's 35 million members, who generate 1.4 billion "claim lines" of data each year, said Marcus Wilson, president of HealthCore, WellPoint's medical outcomes research subsidiary, which WellPoint acquired in 2003 and is overseeing the new project.

"Claims data is the root, but there is also pharmaceutical data from prescriptions that patients fill, and other robust data, like lab results data, that can help us analyze safety issues," he said, including whether a certain drug causes kidney toxicity in patients.

"Drugs are not safe. There is no such thing as a safe drug," he said. "There are benefits and risks" for all medications. However, the system aims to help understand, for instance, if certain drugs pose higher adverse risks to large populations of patients, particular demographic or ethnic groups, or certain types of individuals, such as those undergoing combinations of different medical treatments.

HealthCore has a team of 65, including technology professionals and medical experts working on the Safety Sentinel project, Wilson said.

"We're doing this to benefit our members, and we also have a keen interest in public health," he said. Eventually, HealthCore might offer services to pharmaceutical companies, the FDA, or other organizations that want to use the Safety Sentinel infrastructure for their analysis, he said.

In a statement from the FDA, officials say the WellPoint system fits into the agency's larger goal of having a network of public-private partnerships to identify and address drug safety issues faster.

Initial tests of the Sentinel system showed that it can detect potential adverse events within months, versus the years it can take for the FDA to identify problems because of heavy reliance on the reporting of incidents by doctors.

While WellPoint's effort will use claims, prescription, and lab data, some of the nation's larger health care providers have efforts under way to use patients' clinical data to detect adverse drug events.

Partners HealthCare, which operates several Boston area hospitals, including Massachusetts General and Brigham and Women's, has at least three such projects under way, said Partners CIO John Glaser.

Among those efforts is a collaborative project by Partners, the FDA, industry group eHealth Initiative, and three pharmaceutical companies -- Johnson & Johnson, Eli Lilly, and Pfizer -- who provided funding of $250,000 each. That project is analyzing data from Partner's 4 million patients' e-health records for possible adverse drug results, such as the percentage of people on cholesterol-lowering statin drugs who develop liver issues, he said. Such data mining and analysis efforts also can test if known problems can be somehow gleaned from data, similar to figuring out a math problem based on already knowing the answer.

These sorts of drug safety analysis also can potentially help medical professionals better weigh the risks and benefits of treatments for specific patients, said Glaser. For instance, some analysis might show that Hispanic patients have a higher risk of a specific adverse reaction to a certain drug versus black patients, which could help doctors decide what treatment is safest to prescribe an individual.

"If you blow the alarm, you need to be sure of what you've got" in terms of data providing evidence for certain risks, said Glaser.

As more health care providers deploy e-health record systems to replace paper-based files, the pool of data to help detect public safety issues will grow significantly, said Wilson. And combining those health care provider data-analysis efforts with those by insurance companies and health plans like WellPoint fits together like a puzzle. "When you put these together, you can see things you couldn't see before," he said.

The biggest overall challenge in building these kinds of drug surveillance systems is interoperability issues, said Wilson. That's because data that needs to be analyzed comes from multiple sources, with multiple components, he said. There is no single, national database that will ever contain all claims, lab, pharmaceutical, and other clinical data for all patients, he said. So, the focus is on "how to make it all sync and work quickly," he said.

Glaser said that all the individual efforts under way will eventually help provide a more comprehensive and improved drug safety surveillance system for the public's good.

"If you're the FDA," Glaser said, "you'll want to evolve into a system that uses a dozen of sources of data to get a good geographic spread, good mix of different kinds of patients, a good mix of claims and nonclaims data."